Marriage Market Versus Job Co-Location:

Location Choices of Ph.D.-Trained Singles and Couples

Xirui Zhang

Department of Economics

Syracuse University

426 Eggers Hall

Syracuse, NY 13244-1020

xzhang44@syr.edu

October 28, 2016

Acknowledgments: I am grateful to Stuart Rosenthal, Alfonso Flores-Lagunes, William C.

Horrace, and Perry Singleton for helpful comments. All errors are my own.

Abstract

This paper investigates the degree to which marriage versus job market opportunities affect the

metropolitan location choices of foreign-born Ph.D.-trained workers. A core modeling principle

is that larger metropolitan areas help to alleviate job market co-location problems for high-

skilled couples while large local own-ethnic populations offer better dating opportunities for

high-skilled foreign-born singles. Using US census data for 1990, 2000 and 2010, I investigate

tradeoffs between dating and job market opportunities using differencing strategies to compare

the location choices of foreign-born versus domestic-born Ph.D.-trained workers.

Single Chinese and Indian PhDs are disproportionately drawn to MSAs with large own-ethnic

college-above populations even when such populations are not necessarily situated in large

metropolitan areas. In contrast, foreign-born Ph.D. couples are drawn to large MSAs regardless

of the size of the local own-ethnic population. These and other patterns suggest that single

foreign-born Ph.D. workers are partly willing to forgo the greater labor market opportunities

found in large metropolitan areas in exchange for a more active dating scene. Among married

foreign-born Ph.D. workers, results echo those from Costa and Kahn (2000) that highly educated

couples are disproportionately drawn to large metropolitan areas in order to solve their job

market co-location challenge.

JEL Classification: J1, R2

Key Words: Immigrants, Marriage Markets, Job Colocation, Migration, Ph.D. Workers

1

1. Introduction

From 1990 to 2010, Ph.D. workers increasingly concentrate in large metropolitan areas

(MSAs) in the United States. The share of Ph.D. workers located in large MSAs is 25.2% in

1990, 32.5% in 2000, and 39.4% in 2010 (Appendix Table A1). Moreover, during this period,

nearly half of foreign-born1 Ph.D. workers live in bigger cities that have a large population of

their own ethnicity. One possible driver for such trend is the job co-location pressure faced by

dual-career households — as highlighted by Costa and Kahn (2000), large cities can offer more

opportunities for both spouses in the family to secure a high-quality job that is also within a

reasonable commute from home. Another contributing reason may be that Ph.D.-trained singles

prefer to live in large cities to access more dating opportunities, as pointed out by Compton and

Pollak (2007). This may be especially true for foreign-born Ph.D. workers. That is because

foreign-born PhDs disproportionately marry individuals with at least a college degree and within

their own ethnicity (Table 1, I will discuss shortly). Moreover, their highly educated own-ethnic

potential dates are often not ubiquitous in the host country. For these reasons, single Ph.D.

workers facing thin marriage markets may choose to live in bigger cities that can provide an

active dating environment.2 This paper explores evidence and implications of the relative

importance of marriage markets versus job co-location effects in driving location choices of

Ph.D. workers, in particular for those not born in the US.

The degree to which local marriage markets can attract Ph.D.-trained workers could have

substantial impacts on the intensity of innovative activity in the local areas. In part, that is

because Ph.D.-trained workers, and especially those foreign-born, account for a significant share

1 Foreign-born is defined as those who are not born in the United States and neither parent is American. 2 For example, as indicated by Table 2-1 and 2-1, among the Top 25 MSAs in 2010 that have scores of new-arrival

Ph.D. workers from China and India, most of them are large or midsize cities with historically sizable Chinese and

Indian population.

2

of new product innovation in the United States. According to the RIETI Georgia Tech inventor

survey, 46 % of the inventors who file patents in the U.S. have a doctoral degree (Walsh and

Nagaoka, 2009). Moreover, foreign-born Ph.D. workers are overrepresented in the STEM

(Science, Technology, Engineering, and Mathematics) fields. Among the scientists and engineers

with a Ph.D., nearly 55% are foreign born, and roughly 30% are born in China or India.3 As

Figure 1 shows, across top 30 MSAs that generate a large number of high-quality patents in the

year 2001,4 the number of patents created almost lines up with the MSA count of new-arrival

Ph.D. STEM workers from China and India in 2000.

It is also important to recognize that foreign-born Ph.D. workers face with thin marriage

markets in the sense that there are a few MSAs that can provide them an active dating

environment. That is because Ph.D. workers tend to marry individuals with a college or higher

level of education. Besides, foreign-born Ph.D. workers, especially for those from China and

India, marry primarily within their ethnicity. Table 1 provides evidence of these marriage

patterns. In Panel A Column (6), 71.5% of US-born male Ph.D. holders marry a US-born woman

with at least a four-year college education attainment. Similarly, among Chinese and Indian male

PhDs, 80.6% and 72.4% marry an own-ethnic woman with at least a college degree, respectively.

Moreover, Chinese and Indian female PhDs display a further strong tendency to marry a Ph.D. of

their ethnicity: roughly 45% marry a Ph.D. of own ethnicity compared to just 13% for their male

counterparts. Since there are few MSAs in the U.S. with scores of single, foreign-born college-

educated workers, single foreign-born PhDs may have more restriction in choosing where to live

3 Since 1990, there is an exceptional growth of U.S. patents created by ethnic inventors. According to Kerr (2010),

the share of patents created by Chinese and Indian inventors increased from about 2% in the year 1990 to 9% and to

6% in 2004, respectively. 4 High-quality patents here refer to top cited patents that received a number of citations among top 10% within its

own technology category.

3

compared to their domestic born counterparts, who can find sizable numbers of potential dates in

many more MSAs.5

I focus on and differentiate between two household types when modeling the manner in

which Ph.D. workers choose among different city sizes. One household type is “Super-Power

Singles” (SPS), which refers to Ph.D. workers who reported their marital status as never married,

divorced or widowed. The other household type is “Super-Power Couples” (SPC), in which both

spouses in the household have a Ph.D. degree. The increasing concentration of Ph.D. workers in

large cities may be caused by increasing returns to education, rising valuation on urban amenities

such as access to museums and other cultural activities. Moreover, as mentioned earlier, Super-

Power Singles may choose to live in large cities to access an active dating environment, while

Super-Power Couples may prefer large cities to solve their job co-location problem.

To identify the effect of local marriage markets in driving Ph.D. workers’ location

choices, I adopt the triple-differencing strategy of Costa and Kahn (2000). To examine the co-

location pressure, Costa and Kahn (2000) compare power couples (college-educated couples)

with low-power couples (households in which neither spouse has a college education) to

difference out the marriage market effects for singles. This paper highlights the differential

marriage market opportunities between foreign-born versus domestic-born. In particular, I use

Chinese and Indian as representative ethnic groups and compare their trends of location choices

to that of the domestic-born. Such comparison is conducted first within and then across the

household type. The aim of doing so is to rule out the impact of increasing returns to education,

different tastes on urban amenities, and different pressure for job co-location. The procedure and

the underlying assumptions for these purposes are made clear later in the paper.

5 For more details, see Appendix Table A2-1 and A2-2, which document the distribution of college above workers

among U.S. cities by ethnicity.

4

To better capture the size and quality of labor and marriage markets, I estimate a three-

tiered model in which the triple-differencing strategy is conducted for different classifications of

city size categories. The first-tier model, following Costa and Kahn (2000), uses the MSA’s total

population to classify MSAs into large, middle and small size. In contrast, the second-tier model

uses the size of the own-ethnic population present in the MSA to classify city size categories.

The third-tier model combines the criteria applied in the first two models: large and midsize

MSAs in terms of the total population are further classified into those with a large or small size

of the own-ethnic college-above population. The third-tier model is the preferred one because it

controls for the scale and quality of both Ph.D. labor and marriage markets in a given city type,

which also to some extent capture the density of potential dates in the local area. All of these

estimations use US census data for the survey years 1990, 2000 and 2010.

The paper has two main findings as follows. First, facing thinner marriage markets, the

stronger desire to access an active dating environment causes immigrant Ph.D.-trained singles to

live in cities with denser potential dates. Foreign-born single Ph.D. workers on average display a

stronger tendency relative to their US-born counterparts to live in MSAs with a large size of their

own-ethnic population. In particular, the stronger marriage market effect explains the greater

tendency of Chinese and Indian Ph.D. women to live in midsize MSAs with a large size of the

own-ethnic college-above population by 4 to 7 percentage points. Evidence on the marriage

market effect is present for Ph.D. men but less robust. This gender disparity is consistent with the

fact that foreign-born Ph.D. women face with thinner marriage markets as compared to the Ph.D.

men within their own ethnicity, for it is more common for men to marry less educated women

than the reverse situation. Secondly, findings in this paper suggest that the size and quality of

local labor markets are key drivers of the location choices of Super-Power Couples. Chinese and

5

Indian Super-Power Couples display a stronger tendency relative to their US-born counterparts

to live in large MSAs in terms of total population, regardless of the size of their own ethnic

population in the local area. Job co-location effects seem to dominate marriage market effects in

explaining the greater increasing tendency of Chinese and Indian Ph.D. workers to live in large

MSAs.

The rest of the paper is organized as follows. Section 2 reviews previous studies and

highlights the contributions of this paper. Section 3 uses a multinomial logit model to analyze the

Ph.D. workers’ location choices and presents the predicted average probabilities of choosing

among different city sizes for the survey years 1990, 2000 and 2010. Section 4 describes the

differencing strategy to investigate the relative importance of marriage market and job co-

location effects. Section 5 first presents key results of the three-tiered model based on different

classifications of city size categories, and then a robustness check is conducted by estimating a

linear probability model for a more restricted subsample. Section 6 discusses the policy

implication of this paper and Section 7 concludes.

2. Previous Studies

Highly trained workers tend to move to places with high-quality jobs and are sensitive to

regional wage differences (e.g. Borjas, 2001;6 Chen and Rosenthal, 2008;7 Dahl and Sorenson,

2010). On the other hand, highly educated workers often place high weights on social factors and

urban amenities when choosing a metropolitan location. Dahl and Sorenson (2010) use panel

6 Borjas (2001) find that the new-arrival immigrants, compared to their US-born counterparts, are much more likely

to be clustered in states that offer the highest wages for the types of skills they can offer. Fang and Brown (1999)

also find that Chinese highly educated workers are not closely tied to the ethnic enclave economy as their low-

skilled peers do. 7 Chen and Rosenthal (2008) demonstrate that young, highly educated households tend to move to places with high

quality of business environment and that this tendency is especially pronounced among power couples who are

subjected to job market co-location problem.

6

data of the Danish population to show that technical workers have strong preferences for living

close to family and friends.8 Gottlieb and Joseph (2006) estimate random parameter logit models

to investigate the migration decisions of science and technology graduates across the U.S. cities.

They find that Ph.D. graduates pay greater attention to location-specific amenities than other

degree holders and that some foreign-born students after graduation migrate to places where their

ethnic groups are concentrated.

Marriage market opportunities are an important location-specific amenity that might

affect the migration decisions of highly educated singles, but there is very limited work

investigating this topic. Gautier et al. (2010) provide evidence that singles are willing to pay a

housing price premium to locate in cities, the dense areas where they can meet more potential

partners than in rural areas.9 In this paper, I focus on the impact of marriage market opportunities

in driving the metropolitan location choices for Ph.D.-trained singles.

In particular, to highlight the impact of the local marriage market on foreign-born versus

domestic born Ph.D. workers, I modify and develop the differencing strategy used in Costa and

Kahn (2000). Costa and Kahn (2000) put forward job co-location pressure in explaining the

increasing concentration of college-educated couples (“Power Couples”) in large MSAs from

1940 to 1990. As mentioned earlier, the key idea is that large MSAs appeal to power couples

because such cities can provide high-quality jobs for both spouses in the dual career household.

To identify such a job co-location effect, Costa and Kahn (2000) compare the location choice of

8 Previous studies (e.g., Altonji and Card, 1991; Card, 2001; Peri 2011; Hunt and Gauthier-Loiselle, 2010; and Kerr

and Lincoln, 2010) have used this phenomenon to examine the impact of immigrants on the local economy in the

host country. They usually use the interaction of past immigrant stocks and migration trends as an instrument for

observed local changes. 9 This is consistent with the literature on search and matching (e.g. see discussion in Burdett and Coles, 1999) that

emphasized that the long-term partnership formation as with marriage and employment is a time-consuming activity

because of market frictions and heterogeneity of agents. Therefore, access to a thick local market increases the

arrival rate and contact with potential partners.

7

power couples with that of low-power couples (households in which neither spouse has a college

education), and difference out the marriage market effects for singles. Their work suggests that

large cities will continue to attract highly educated couples, and such migration will contribute to

the matching in labor markets as well as the city growth. Building upon their work, this paper

highlights the differential marriage market effects for foreign-born versus US-born. Moreover, I

focus on the migration of Ph.D.-trained individuals instead of the college-educated workers. If

the local marriage market does affect the location choices of new-arrival foreign-born PhDs, it

will provide an important margin to adjust the innovative activities across U.S. cities.10

Especially for struggling cities like Cleveland, in order to reinvent cities and achieve urban

growth, it would be appealing for policymakers to invite the foreign-born Ph.D. workers to local

areas by using some policy instruments.

3. Location Choices of Ph.D. Workers

This section first models how Ph.D. workers choose among different city size categories.

Then a multinomial logit model is estimated to predict Ph.D. workers’ probabilities of living in

large, middle and small size MSAs for the survey years 1990, 2000, and 2010, conditional on

their gender, ethnicity (US-born, Chinese or Indian), and household type (Super-Power Couples

or Super-Power Singles). At the end of this section, the year-specific tendencies of location

choices are documented for each gender-ethnicity-household group. All of the estimations are

10 Kerr (2010) points outs that spatial adjustments of innovation are highly associated with the mobility of the

scientists and engineers that develop the technology. Kerr (2010) finds that patenting migrates faster for

technologies that employ immigrant inventors intensively.

8

conducted using U.S. Census data from the Integrated Public-Use Microdata Series (IPUMS,

Ruggles et al., 2010).11

Consider a Ph.D.-trained individual i has preferences over available city size s (e.g., s can

be chosen from small MSAs, midsize MSAs, and large MSAs). It can be described by a utility

function as follows:

𝑈𝑖𝑠 = 𝛽′𝑋𝑖 + 휀𝑖𝑠, (1)

where 𝑋𝑖 denotes a vector of individual specific characteristics (e.g., age, age squared, English

speaking ability, in a STEM occupation or not, etc.), 𝛽 is the coefficient, and error term 휀𝑖𝑠

allows for different evaluations on living in city size s.

Individuals choose to live in the city size s that maximizes their expected utilities. If the

errors term 휀𝑖𝑠 is independent and identically distributed draws from a Type I extreme value

distribution, then the probability that individual i chooses city size s is:

𝑃(𝑌𝑖 = 𝑠) = exp{𝛽′𝑋𝑖 + 휀𝑖𝑠} ∑ exp{𝛽′𝑋𝑖 + 휀𝑖𝑠}𝑠⁄ . (2)

Doing some transformation on equation (2) yields the estimating equation as follows:

log(𝑃𝑠 𝑃𝑠𝑚𝑎𝑙𝑙⁄ ) = β′𝑋𝑖 +휀𝑖𝑠, (3)

where 𝑃𝑠𝑚𝑎𝑙𝑙 is the probability that individual i chooses to live in a small MSA, which has been

set up as the baseline group in the multinomial logit model.

Then I introduce how to estimate equation (3) to obtain the average tendency to live in

city size s for each gender-ethnicity-household group: Probability (Live in city size s | gender,

ethnicity, household type). For each survey year t (1990, 2000 or 2010), conditional on gender

and household type, I first run equation (3) separately for foreign-born and US-born to get the

11 Data for 1990 and 2000 are based on 5 percent samples of the underlying Census population while data for 2010

are from the 1 percent sample from American Community Surveys (ACS).

9

coefficient estimates �̂�. Then I fix the age at 35 years old and calculate the average standardized

probabilities of living in city size s in year t, denoted as �̂�𝑠,𝑡, for Ph.D. workers in a specific

gender-ethnicity-household group.

Take single, US-born male Ph.D. workers as an example to illustrate the procedure. To

calculate their propensity to live in large, midsize, and small MSAs in 2010, denoted as

�̂�𝑙𝑎𝑟𝑔𝑒,2010,�̂�𝑚𝑖𝑑𝑠𝑖𝑧𝑒,2010, and �̂�𝑠𝑚𝑎𝑙𝑙,2010, I use a sample consisting of 9,203 US-born single male

workers that had a Ph.D. or master equivalent education level and aged between 25 and 50.

Equation (3) is estimated in which I control for age, age squared, English language ability, in a

STEM occupation or not, self-employed or not, Ph.D. holder or not, race is white or not, and

whether this household is linguistically isolated.12 Once the coefficient estimates �̂� are obtained,

I fix the age at 35 and calculate the average standardized probabilities among all the single, US-

born male Ph.D. workers whose race is white. It yields the propensities to live in different city

sizes: �̂�𝑙𝑎𝑟𝑔𝑒,2010,�̂�𝑚𝑖𝑑𝑠𝑖𝑧𝑒,2010, and �̂�𝑠𝑚𝑎𝑙𝑙,2010.

Table 3-1 and 3-2 report Ph.D. workers’ tendency to living in large, midsize, and small

MSAs in 1990, 2000 and 2010, where the city size categories are classified based on the MSA’s

total population for those aged 15 above.13 Table 3-1 is for single Ph.D. workers (Super-Power

Singles) and Table 3-2 for married Ph.D. workers who have an own-ethnic Ph.D.-holding spouse

(Super-Power Couples). It should be noted here that all Ph.D. workers, regardless of their

gender, ethnicity, and household type, increasingly concentrate in large MSAs from 1990 to

2010. Moreover, for each survey year, conditional on gender and household type, Chinese and

12 Households are linguistically isolated if either no person age 14+ speaks only English at home, or no person age

14+ who speaks a language other than English at home speaks English "very well". 13 Following Costa and Kahn (2000), I define large MSAs as metropolitan areas in which the total population for

those who are aged 15 above was at least 2 million, midsize MSAs as those between 2 million and 250,000, and

small MSAs as those less than 250,000 or in rural areas. For other classifications of city size categories in the second

and third tier models, see the yearly tendency of location choices in Appendix 4-1, 4-2, 5-1 and 5-2.

10

Indian PhDs display a greater tendency to live in large MSAs relative to their US-born

counterparts.

Table 3-1 shows single Ph.D. workers that are from China and India are on average more

likely to be located in large MSAs relative to their US-born counterparts. Their probabilities to

live in large MSAs are 31% to 38% in 1990, 40% to 43% in 2000, and 44% to 50% in 2010,

larger than those of their US-born counterparts — roughly 26% in 1990, 34% in 2000, and 37%

to 41% in 2010. This disparity is also present in Table 3-2 among married Ph.D. workers in

Super-Power Couples (SPC). In 2010, the average probability to live in large MSAs is roughly

50% for Chinese and Indian Super-Power Couples, larger than just 33% for comparable US-born

Super-Power Couples. In particular, married Chinese female Ph.D. workers that have a Chinese

Ph.D. husband display the most increasing tendency to live in large cities: their probability of

being large MSAs increases by roughly 11 percentage points from the year 1990 to 2000, and

then further rises by 12 percentage points from 2000 to 2010.

4. Hypotheses and Identification Strategy

What factors would explain the increasing tendency of Ph.D. workers to live in large

cities? To what extent is such migration trend driven by social factors relative to the concern for

job opportunities? In order to answer these questions, this section first considers the benefits of

being in large MSAs for Ph.D. workers. Then the identification strategy is outlined (Table 4) to

investigate the degree to which the local marriage markets affect the location choices of foreign-

born Ph.D. workers differently compared to their domestic-born counterparts.

Large cities appeal to Ph.D. workers because they have higher returns to education and

better urban amenities such as access to cultural activities. Moreover, for Super-Power Couples

11

(SPC), large cities are more likely to offer high-quality jobs for both spouses in the family to

solve their co-location problems. While for Super-Power Singles (SPS), large cities may provide

better access to an active dating environment as such cities often have more educated young

people within their ethnicity. Following this analysis, Ph.D. workers’ increasing tendency to live

in large cities may be caused by the increasing returns to city size, the rising valuations on urban

amenities, a rising pressure for job co-location, or a strong desire to access a better marriage

market.

As outlined in Table 4, a differencing strategy is adopted to identify the role of local

marriage market in driving the location choices of Ph.D.-trained workers. For simplicity, I use

Chinese as an example here to represent an ethnic group that is not US-born. Briefly, this method

first compares the trend estimates of Chinese Ph.D. workers to live in a given city type to that of

US-born Ph.D. workers within the household type (Row 3 and Row 6), and then the double

differencing estimators obtained are compared across household types (Row 7). The rest of this

section will describe the differencing strategy in details and clarify the underlying assumptions.

The first assumption is as follows. There is no discrimination on immigrants in the Ph.D.

labor markets. Within gender group, returns to education in a given city type are same for all

Ph.D.-trained workers, both foreign-born and domestic-born.

Given the first assumption, within the household type, the effect from returns to

education can be differenced out when the trend estimators of living in a given city size are

compared across ethnicities. For example, in Row 3 for married Ph.D. workers in Super-Power

Couples (SPC), subtracting the trend estimator of US-born SPC (Row 2) from that of Chinese

SPC (Row 1) yields the double-differencing estimator for Super-Power Couples between

Chinese and US-born, (SPC𝐶ℎ − SPC𝑈𝑆). This double-differencing estimator indicates the

12

combined effects of their different valuations on urban amenities (𝐴𝐶ℎ − 𝐴𝑈𝑆), plus their

differential pressure from job co-location (𝐶𝐶ℎ − 𝐶𝑈𝑆). Similarly, in Row 6, the double-

differencing estimator for single Ph.D. workers between Chinese and US-born, (SPS𝐶ℎ −

SPS𝑈𝑆), reveals their different valuations on urban amenities (𝐴𝐶ℎ − 𝐴𝑈𝑆) and local marriage

markets (𝑀𝐶ℎ −𝑀𝑈𝑆).

Once the double-differencing estimators for SPC (Row 3) and SPS (Row 6) are obtained,

I difference them across the household type to generate the triple-differencing estimator, as

indicated in Row 7. In doing so, the underlying Assumption 2 is made as follows. Chinese and

US-born Ph.D. workers may have different tastes for urban amenities, but such disparity in their

valuation on urban amenities (𝐴𝐶ℎ − 𝐴𝑈𝑆) remains the same for both Super-Power Couples and

Super-Power Singles.14 Given this assumption, the impact of different tastes on urban amenities

(𝐴𝐶ℎ − 𝐴𝑈𝑆) can be ruled out when the triple differencing is conducted across these two

household types. As a result, in Table 4 Row 7, the triple-differencing estimator between

Chinese and US-born indicate the difference between two effects: the differential marriage

market effects and the differential job co-location effects, that is, (𝑀𝐶ℎ −𝑀𝑈𝑆) − (𝐶𝐶ℎ − 𝐶𝑈𝑆).

To predict the sign of the triple-differencing estimator, the following discussion analyzes

the signs of two terms, (𝑀𝐶ℎ −𝑀𝑈𝑆) and (𝐶𝐶ℎ − 𝐶𝑈𝑆). First, it should be noted that the

marriage market effect in large cities is expected to be stronger for Chinese than that for their

US-born counterparts, that is, (𝑀𝐶ℎ −𝑀𝑈𝑆) > 0. That is because Chinese and Indian single

PhDs, as shown in the marriage patterns of Table 1, prefer to a find a spouse with a college or

14 This assumption should hold if the valuation on urban amenity for married Ph.D. workers relative to single Ph.D.

workers stays the same across ethnicities. For example, if Chines married PhDs care about the school quality in local

areas as US-born married PhDs do, while single PhDs, both Chinese and US-born, do not have concern for the

school quality, the valuation on school quality should be differenced out.

13

higher level of education and within their ethnicity. Moreover, there are few MSAs in the U.S.

with scores of single, foreign-born college-educated workers but many more MSAs with sizable

numbers of domestic-born college-educated singles. For these reasons, single, foreign-born

Ph.D. workers often face thin marriage markets compared to their US-born counterparts, and

large cities with better marriage markets may be more attractive to them.

Regarding the sign of (𝐶𝐶ℎ − 𝐶𝑈𝑆), it is important to recognize that a given city type

may have different depth of Ph.D. labor markets between foreign-born and domestic-born, and

whether (𝐶𝐶ℎ − 𝐶𝑈𝑆) equals to zero depends on the standard that is used to classify city size

categories. For example, if the MSA’s total population is a good proxy for the size of the Ph.D.

labor market in the local area, regardless of the Ph.D. workers’ ethnicities, (𝐶𝐶ℎ − 𝐶𝑈𝑆) = 0 is

expected to be held when the city size categories are classified based on the MSA’s total

population. In contrast, this condition may not hold if Chinese PhDs are heavily concentrated in

selected industries and occupations as previous studies (Mandorff, 2007) has pointed out, and

these sorts of jobs are only available in a few numbers of large cities.15 In this circumstance,

Chinese Super-Power Couples may put higher weights on large cities to solve their co-location

problems compared to their US-born counterparts; In other words, I expect (𝐶𝐶ℎ − 𝐶𝑈𝑆) > 0.

Given the analysis above, the interpretation of the triple-differencing estimator is

sensitive to the way I classify the city size categories. When (𝐶𝐶ℎ − 𝐶𝑈𝑆) = 0 holds, the triple-

differencing estimator should yield the differential marriage markets effects between Chinese

and US-born Ph.D. workers. For a given city size, the estimator is expected to be positive if

Chinese Ph.D. workers put higher weights on having access to an active local marriage market

15 Another possible reason why large cities are more important for immigrants to solve their colocation problem is

that large cities may have more companies that will effectively sponsor the immigrant Ph.D. workers for H-1B visas

and green cards.

14

than their US-born counterparts. If (𝐶𝐶ℎ − 𝐶𝑈𝑆) > 0 holds, the triple-differencing estimator

should indicate the relative importance of differential marriage market effects over the

differential co-location effects between Chinese and US-born Ph.D. workers. Under this

circumstance, a positive triple-differencing estimator suggests that the marriage market effects

dominate the co-location effects, while a negative one indicates the other way around.

5. Results

This section begins with estimating a three-tiered model in which the triple differencing

strategy outlined in section 4 is conducted under different classifications of city size categories.

The discussion mainly focuses on results of the period from 1990 to 2000, while the results of

the period 1990-2010 are also documented. In particular, from 1990 to 2000, high trained

individuals from China and India dramatically increase in the United States; moreover, the

number of new-arrival Chinese and Indian PhDs that work in the STEM fields more than

doubled.16 In the first-tier model, the MSA’s total population for those aged 15 above is used to

classify MSAs into large, middle and small sizes. The second-tier model uses the size of the

own-ethnic population present in the MSA instead of the total population. The third-tier model,

which is the preferred one, combines the criteria applied in the first two models to controls for

the size of Ph.D. labor and marriage markets in a given city type, and also, to some extent,

capture the density of potential dates in the local area.

16 For example, in 1990, the number of new-arrival Chinese Ph.D. workers that have been in U.S. no more than ten

years is 3,818, and this figure increases by more than four-folds to 20,905 in 2000 and reaches 33,111 in 2010.The

number of new-arrival Indian STEM Ph.D. workers is 3,711 in 1990, 9,641 in 2000, and 14,937 in 2010. Person

weights are used to calculate these numbers. Regarding the total count of Chinese STEM Ph.D. workers in the U.S.,

it is 13,169 in 1990, 45,873 in 2000, and 81,972 in 2010. The total number of Indian STEM Ph.D. workers is 10,106

in 1990, 24,868 in 2000, and 39,947 in 2010.

15

One concern for the triple differencing approach is that some Ph.D. workers chose to live

in an MSA with an active dating scene when they were single, and they may stay in the same

MSA after they have been married. In this case, the triple differencing approach may

underestimate the marriage market effect, for the concentration of these sorts of couples should

be attributable to the marriage market effect but it has been mistreated as job collocation effect.

To address this concern, I conduct a robustness check by estimating a linear probability model

for a sample of young Ph.D. workers that include both marriage and single. The goal of this

check is to compare the migration decisions of young single Ph.D. workers to those comparable

married Ph.D. workers who made their moving decisions after they have being married.

Moreover, observations in this sample also include other foreign-born Ph.D. workers beyond

Chinese and Indian PhDs. Briefly summarizing the findings, improving the local dating

environment decrease the probability of single foreign-born PhDs to move while no such effect

arises for married Ph.D. workers or domestic-born singles.

5.1. City Size Categories Based On the Total Population

Following Cost and Kahn (2000), the first-tier model categorizes MSAs into large,

middle, and small size based on the total population size. Table 5-1 shows Ph.D. workers’ trend

in the propensity to live in large, middle and small size MSAs from 1990 to 2000. For each

gender-ethnicity-household group, such first-differencing estimates are obtained by differencing

their tendencies of location choices across two survey years 2000 and 1990, which have been

documented in Table 3-1 and 3-2. Table 5-2 reports the double-differencing estimates — the

differential trends between Chinese (or Indian) and US-born from 1990 to 2000, conditional on

gender and household type. Table 5-3 shows the triple-differencing estimates by gender.

16

Table 5-1 reveals that all Ph.D. workers, regardless of their gender, ethnicity, and

household type, display an increasing tendency to live in large MSAs from the year 1990 to

2000. Among single Ph.D. workers in Column 1 and 3, the trend estimates of the propensity to

live in the large MSAs range from 0.06 to 0.09. Such increasing tendency of Super-Power

Singles (SPS) to being in large MSAs may be caused by the increasing returns to Ph.D.

education, the rising valuation on urban amenities, and the appealing local marriage markets that

large MSAs can provide. Among Super-Power Couples (SPC), as indicated by Column 2, the

probabilities of being in large MSAs increases by roughly 0.04 for US-born married Ph.D.

women, smaller than those of Chinese and Indian women, 0.11 and 0.06, respectively. Such

disparity is also present among married male PhDs in Super-Power Couples in Column 4, with

smaller magnitudes for the increasing trends across all ethnicities. The increasing tendency for

Super-Power Couples to be located in large cities, as indicated in Section 4, may result from the

increasing returns to Ph.D. education, the rising valuation on urban amenities, and the increasing

pressure for job co-location.

Table 5-2 presents the estimates of the differential trends between foreign-born and US-

born, and different patterns occur for Super-Power Singles versus Super-Power Couples. Among

single Ph.D. workers in both Panel A and B, the trends of the foreign-born relative to the US-

born are positive for living in midsize MSAs. These estimates suggest that from 1990 to 2000,

the Chinese and Indian single Ph.D. workers’ greater valuation on urban amenities and local

marriage markets explain their greater increasing tendency relative to their US-born counterparts

to live in midsize MSAs. While regarding the trend of foreign-born single PhDs to be located in

large MSAs relative to their US-born counterparts, most estimates are non-positive except for the

Indian single male PhDs in Column 3. In contrast, for both male and female married Ph.D.

17

workers in Super-Power Couples, the trends of Chinese and Indian in the propensity to live in

large MSAs is significantly greater than those of their US-born counterparts. These estimates,

compared with those of Super Power Singles, potentially suggest that large cities are especially

attractive to foreign-born Ph.D. couples by enabling them to solve their job co-location

problems.

When further differencing these estimate in Table 5-2 across the household types, the

resulting triple-differencing estimates for the period 1990-2000 are positive for being midsize

MSAs but negative for being large MSAs, as reported in the first two columns of Table 5-3. As

analyzed in Section 4, these estimates indicate that compared to their US-born counterparts, the

impact of local marriage markets for immigrant single Ph.D. workers are relatively weak in large

MSAs but strong in midsize MSAs. Why do the local marriage markets in large MSAs appear to

be less attractive to Chinese and Indian single Ph.D. workers relative to their US-born

counterparts? One possible explanation is that the first–tier model uses the total population to

categorize city sizes, but this measure may not be good enough to capture the thickness of local

marriage markets for immigrant single Ph.D. workers — because foreign-born Ph.D. workers

disproportionately marry within their own ethnicity, as indicated by marriage patterns in Table 1.

Also, it should be noted that the triple-differencing estimates in Table 5-3 may reveal the

differential pressure for job co-location between the foreign-born and US-born Super-Power

Couples. As mentioned in Section 4, foreign-born Ph.D. workers often concentrate in selected

industries and occupations, which cause them to have more restriction in choosing where to live.

Large MSAs, usually with a mix of diverse industries, provide more high-quality job

opportunities for these foreign-born Super-Power Couples that face thinner labor markets. This

argument is also consistent with the double-differencing estimates in Table 5-2. For these

18

reasons, the negative triple-differencing estimates for being large MSAs may suggest that the

impact of job co-location dominates the effect of the marriage market in explaining the greater

increasing trend of Chinese and Indian Ph.D. workers to live in large cities relative to their US-

born counterparts.

5.2. City Size Categories Based On the Size of the Own-Ethnic Population

This section estimates the second-tier model in which for each person, all MSAs are

reclassified into three city size categories based on the own-ethnic population for those aged 15

above and present in that MSA.17 Compared to the first-tier model, this model does a better job

in capturing the size of the local marriage market for both foreign-born and US-born PhDs. On

the other hand, the city size in terms of the own-ethnic population does not necessarily control

for the size of Ph.D. labor markets in the local areas. Therefore, a differential effect of job co-

location may be present for a given city size between foreign-born and US-born Super-Power

Couples. For these reasons, the triple-differencing estimates of the second-tier model in Table 6,

as indicated earlier in Section 4, potentially reveal the relative importance of the marriage market

effect versus the pressure for job co-location in driving the Ph.D. workers’ location choices.

Specifically, city size categories in the second-tier model are defined as follows. For a

given person, large own-ethnic MSAs are metropolitan areas in which his/her own ethnic

population rank within top 15 among all U.S. MSAs, midsize own-ethnic MSAs as those that

rank outside top 15 but have at least 1,000 people within the ethnicity of this person (the cutoff

number is 250,000 for the US-born) present in the local areas, and the rest of MSAs are all

classified as small own-ethnic MSAs. Based on such classification of city size categories,

17 Table A6-1 and A6-2 show the predicted probabilities of being in different city sizes for Super-Power Singles and

Super-Power Couples, respectively.

19

Appendix Table A4-1 and A4-2 document the year-specific location choices of Super-Power

Singles and Super-Power Couples, respectively.

As shown in Appendix Table A4-1 and A4-2, Chinese and Indian Ph.D. workers, both

single and married, are more likely to concentrate in MSAs that have a large size of the own-

ethnic population compared to their US-born counterparts. Moreover, such disparity is increasing

from 1990 to 2000. For example, the average probabilities of living in large own-ethnic MSAs

increase by 0.08 — from 0.53 to 0.61— for Chinese single female PhDs, while the

corresponding probabilities stay at 0.32 for their US-born counterparts.18 Such discrepancy in

location trend is also present among Super Power Couples in Appendix Table 4-2. In particular,

during the period of 1990 to 2000, the probabilities of being in large own-ethnic MSAs

significantly increase by roughly 0.06 and 0.09 for Chinese and Indian Super-Power Couples,

respectively, while it decreases by approximately 0.02 for US-born Super-Power Couples.

Table 6 presents the triple-differencing estimates of the second-tier model, which provide

evidence that marriage market opportunities are the driving force that causes Ph.D. women to

live in cities with a large population of their own ethnicity. Column 1 shows that from 1990 to

2000, the marriage market effect dominates the job co-location effect in explaining Indian Ph.D.

women’s greater increasing tendency to live in large own-ethnic MSAs relative to their US-born

counterparts, and the magnitude is approximately 3 percentage points. While for Chinese Ph.D.

women relative to their US-born counterparts in Panel A, the estimate is positive but not

statistically different from zero. Conversely, among male Ph.D. workers in Table 6 Column 2,

the triple-differencing estimates for the period 1990-2000 are negative for living in large own-

18 Such disparity in geographic sorting also exists among single male Ph.D. workers. The estimate for the tendency

to live in large own-ethnic MSAs stays at roughly 0.49 for Chinese single men during the 1990-2000 period, and it

increases from 0.39 to 0.47 for Indian single men, but decreases from 0.32 to 0.30 for US-born single men.

20

ethnic MSAs. It suggests that in explaining foreign-born Ph.D. men’s greater increasing

tendency to live in large own-ethnic MSAs relative to those comparable US-born Ph.D. men, the

impact of the differential concerns for job co-location is larger than that of the differential

marriage markets, by roughly 0.06 and 0.03 for Chinese and Indian, respectively. The disparity

between Ph.D.-trained men and women seems to be consistent with the traditional view that

men, in general, may be more likely to put job opportunities into priority concern than women.

5.3.City Size Categories Considering the Density of Potential Dates

This section presents results for the third-tier model in which city sizes are classified

based on the size of the total population aged 15 above and the size of the own-ethnic population

for those that have at least a four-year college education level (OECoAbv) in the MSA. As

mentioned earlier, this model is the preferred one because it takes into account the size of the

local Ph.D. labor and marriage markets. Moreover, controlling for these two populations – the

total population and the own-ethnic highly-educated population – at the same time will shed light

on the density of potential dates in the local area, which may provide a better proxy for the

quality of the local dating environment. For these reasons, based on such city size classification,

the job co-location effect in a given city type is expected to be the same for both foreign-born

and US-born Super-Power Couples. As indicated in Section 4, the triple-differencing estimates in

Table 7-3 can be interpreted as the differential marriage market effects between foreign-born and

domestic-born Ph.D. workers. The discussion below first introduces the city size classification in

details and the trends of Ph.D. workers’ location choices are documented in Table 7-1 and 7-2,

and then Table 7-3 shows the triple-differencing estimates of the third-tier model.

21

The third-tier model first classifies MSAs into large, middle, and small sizes based on the

size of the total population as the first-tier model does, and the baseline group is still small

MSAs. Meanwhile, for each person, large and midsize MSAs are further categorized into those

with or without a large size of the own-ethnic population whose education level is at least a four-

year college (OECoAbv). A large or midsize MSA is defined as having large OECoAbv if the

size of the own-ethnic college-above population present in that MSA rank within top 15 among

all U.S. MSAs in a given year; otherwise that MSA is classified as having small OECoAbv. For

example, in the year 1990, Boston, MA is a large MSA with large OECoAbv for Chinese

workers, and a large MSA with small OECoAbv for Indian workers. In contrast, Detroit, MI in

1990 is a large MSA with large OECoAbv for Indians but with small OECoAbv for Chinese.

Table 7-1 and 7-2 present the trend estimates of location choices for single and married

Ph.D. workers, respectively.19 Table 7-1 shows that from 1990 to 2000, single PhDs generally

display an increasing tendency to live in large MSAs. Moreover, Chinese and Indian single Ph.D.

workers display a further strong tendency to live in midsize cities that have a relatively high

density of potential dates. During this period, the probabilities of living in midsize MSAs with a

large size of the own-ethnic college-above population (OECoAbv) increase by 0.01 to 0.03 for

Chinese and Indian single PhDs and decrease by 0.02 to 0.03 for their US-born counterparts.

Conversely, among Super-Power Couples in Table 7-2, only Indian married Ph.D. men

increasingly concentrate in midsize MSAs with large OECoAbv. Instead, all Super-Power

Couples increasingly concentrate in large MSAs, regardless of the size of the local potential

dating pool. These estimates may suggest that the size of the total population play a more

19 Appendix Table A5-1 and A5-2, for single and married Ph.D. workers, respectively, document the year-specific

average probabilities of choosing among these five city size categories.

22

important role than the size of the own-ethnic population to capture the quality of the local labor

market for Super-Power Couples.

The discussion below focuses on the triple-differencing estimates of the third-tier model

presented in Table 7-3. First, consider the estimates of being in midsize MSAs with large own-

ethnic college-above population – a city size category that probably has the most denser potential

dates. In Table 7-3 Column 1, the estimates for Chinese and Indian Ph.D. women relative to US-

born Ph.D. women are roughly 0.07 (Panel A) and 0.04 (Panel B), respectively. These triple-

differencing estimates indicate that from 1990 to 2000, the stronger marriage market effect for

Chinese and Indian Ph.D. women explains their greater increasing tendency to live in midsize

MSAs that also have a large size of the own-ethnic college-above population. Such effect is still

present for Chinese Ph.D. men in Column 2, but with a smaller magnitude 0.02, while for Indian

Ph.D. men, the estimate turns out to be negative. This gender disparity seems to be consistent

with the thinner marriage markets faced by immigrant Ph.D. women as compared to Ph.D. men

of their ethnicity, for it is more common for men to marry less educated women than the reverse

situation.

Secondly, we look at the estimate of living in large MSAs with a large size of the own-

ethnic college-above population. From 1990 to 2000, the triple-differencing estimates are

positive for Indian PhDs (Panel B) but negative for Chinese PhDs (Panel A). It suggests that the

stronger marriage market effect for Indian Ph.D. workers relative to their US-born counterparts

explain their greater tendency to live in these areas by approximately 0.01 to 0.02, while it is the

other way around for Chinese Ph.D. workers relative to their US-born counterparts. Comparing

these estimates to those analyzed in the last paragraph, it suggests that regarding the

23

opportunities for foreign-born Ph.D.-trained singles to find an ideal spouse, the absolute number

of potential dates matters, but not as important as the density of potential dates in a local area.

The importance of the density of the local marriage market is also evident among triple-

differencing estimates for the period 1990-2010 in Column 3 and 4 of Table 7-3. Compared to

those of the period 1990-2000, the estimates of a longer horizon reveals that the smaller cities

with denser potential dates present in local areas have increasingly become feasible marriage

markets for immigrant Ph.D. workers. Regarding choosing midsize MSAs with a small size of

own-ethnic college-above population, most estimates in Column 3 and 4 are 0.07 to 0.19 (except

for Indian Ph.D. women), much larger as compared to those in Column 1 and 2. Moreover, for

Ph.D. women to live in small MSAs, the triple-differencing estimates of the period 1990-2000

are not statistically different from zero, while they become significantly positive in the longer

time horizon. These changes are consistent with the massive inflow of highly educated Chinese

and Indian across U.S. MSAs from 2000 to 2010, which makes marriage markets more

geographically expansive for Chinese and Indian Ph.D. workers.20 For example, during this

decade, the mean of Indian college-above population in small MSAs rises from 377 to 801, and

in midsize MSAs from 1,986 to 3,226. The new emerging marriage markets in smaller cities may

have a potential to attract Chinese and Indian Ph.D. workers to these areas beyond the job

opportunities.

20 During this period, the total count of college-above workers nearly doubles for Indian group, from 566,533 to

1,055,009, and for Chinese group this number increases by nearly 50% from 465,949 to 671,514. This large inflow

increases the size as well as the density of potential dates for the Chinese and Indian Ph.D. workers at smaller cities.

24

5.4. Robustness Check

This section further compares the migration decision of single to that of married for

young Ph.D. workers by estimating a linear probability model. The previous triple differencing

approach, as pointed out by Compton and Pollak (2007), may have underestimated the marriage

market effect because some single Ph.D. workers chose to live in an MSA with an active dating

scene, and they stay in the same MSA after marriage.21 This location lock will tend to occur

when moving costs are higher than the benefits from relocating the household. To address this

concern, I restrict married PhDs to those young (age 25 to 44), full-time workers who made their

moving decisions after they have being married, and I pool this subsample with young, single

Ph.D. workers to conduct a comparison. The key hypothesis I test here is that increasing the size

of the local potential dating pool will reduce the probability of single Ph.D. workers, especially

for those foreign-born, to move to another MSA, while no such effect is expected to arise for

married Ph.D. workers.

The linear probability model is as follows.

𝑀𝑜𝑣𝑒𝑖 =𝛼1𝑀𝑒,𝑐,𝑡 + 𝛼2𝐹𝑜𝑟𝑒𝑖𝑔𝑛𝐵𝑜𝑟𝑛 + 𝛼3𝑆𝑖𝑛𝑔𝑙𝑒 + 𝛼4𝑀𝑒,𝑐,𝑡 × 𝑆𝑖𝑛𝑔𝑙𝑒 +

𝛼5𝑀𝑒,𝑐,𝑡 × 𝑆𝑖𝑛𝑔𝑙𝑒 × 𝐹𝑜𝑟𝑒𝑖𝑔𝑛𝐵𝑜𝑟𝑛 + 𝑋𝑖𝛽 + 𝐴𝑐,𝑡𝛾 + 𝛿𝑐 + 𝜂𝑗 + 𝜇𝑡 + ԑ𝑖, (4)

where i denotes individual, j is for industry, e for ethnic group, c for the previous

residential MSA, and t for year. The dependent variable is whether a given Ph.D. worker i moved

to a different MSA between year t and t+1: coded as one if that person moved and zero if stayed

in the same MSA. The key regressor is individual i’s local marriage market, denoted as 𝑀𝑒,𝑐,𝑡,

which is measured by the number of singles of the opposite gender with a college or higher level

21 Compton and Pollak (2007) revisit the job market colocation effect put forward by Costa and Kahn (2000), and

they argue that the rising concentration of highly educated couples in large cities has more to do with the greater

opportunity for highly educated singles to meet in large urban centers.

25

of education and within individual i’s ethnic group e, at MSA c, for year t. The coefficient of

interest is 𝛼5 on the triple interaction term — 𝑀𝑒,𝑐,𝑡 interacted with the indicator of being foreign

born and the indicator of being single in marital status. As discussed earlier, the sign of 𝛼5 is

expected to be negative.

The estimating equation controls for a vector of individual i’s demographic and

socioeconomic attributes 𝑋𝑖,22 as well as location-specific attributes 𝐴𝑐,𝑡 of the previous

residential MSA c for year t.23 The estimating equation also includes MSA fixed effects 𝛿𝑐 to

capture the time-invariant location attributes like sunshine, river, mountains, air quality, distance

to the country border, coastal proximity, the number of universities, and the cost of living, etc.

Industry fixed effects 𝜂𝑗 is also included to capture the time-invariant attributes for a specific

industry such as its ethnic concentration, gender density, average skill levels. Year fixed effects

𝜇𝑡 control for the national-wide shock in a particular year and ԑ𝑖 is the classical error term.

Table 8 presents results for this linear probability model. In Column 1 among young

female Ph.D. workers, locating in the previous MSA with 1,000 more single men of their own

ethnicity and with a college or higher level of education — the likely dating pool — decreases

the probability of foreign-born single women to move to another MSA by 3.3 percentage point.

While no significant impact arises for married Ph.D. women, nor for single Ph.D. women that are

domestic-born. Male Ph.D. workers in Column 2 obtain similar estimates. Increasing the size of

22 For example, 𝑋𝑖 includes age, age squared, having own children at home, English speaking ability, occupation's

median earned income, residing in the U.S. no more than ten years, in a STEM occupation, Hispanic origin, black

race, living in the home state. In particular, I use the variable (“ERSCOR90”) constructed by IPUMS to measure

occupation's median earned income. This variable reports the percentage of persons in occupations having lower

standardized median earnings than the respondent's occupation. For more details, see https://usa.ipums.org/usa-

action/variables/ERSCOR90#description_section. 23 To capture the job opportunities at the MSA c for year t, I control for the total population, the number of workers

in own industry j, and average working age and average earnings among all workers in the MSA. To rule out the

effect from being close to family and friends, I also include the number of own-ethnic workers that are present in the

MSA c but are not the potential dates for individual i.

26

the local marriage market has no significant impact for married Ph.D. men, nor for US-born

single Ph.D. men, while it reduces the probability of moving for single foreign-born Ph.D. men

by 3.3 percent. As the first two columns show, the presence of children at home reduces the

Ph.D. workers’ possibilities of moving by roughly 8 percentage points. Since married PhDs, as

the summary statistics suggest, are more likely to have children at home than comparable single

PhDs, Column 3 and 4 further restrict the sample to households with no child at home to rule out

the influence of children. As shown in the last two columns of Table 8, among no-kid

households, the marriage market effect grows to 3.8 percentage points for single foreign-born

Ph.D. women while it remains at 3.4 percentage points for single foreign-born Ph.D. men. Again,

no such effect is present for other demographic groups in the last two columns.

6. Policy Implications

This paper has implications for the geographic distribution of Ph.D. workers across U.S.

cities, which has a potential to affect the innovation productivity and economic growth in local

areas. First, it implies that large cities will continuingly appeal to Ph.D.-trained couples by

providing better opportunities for both spouses in the household to find a high-quality job that is

also within a reasonable commute from their home. It is worth noting that the number of Super-

Power Couples is highly likely to grow in the STEM fields. In part, that is because more Ph.D.-

trained women enter the STEM fields. The female share of Ph.D. STEM workers rises from 24%

in 1990 to 31% in 2000 and 40% in 2010. Moreover, Ph.D. women, display a strong tendency to

find a Ph.D.-holding spouse.24 The resulting growth of Super-Power Couples and severity of job

24 Foreign-born Ph.D. workers are highly likely to form Super-Power Couples (SPC) within their own ethnic group.

Beyond their strong tendency to marry a highly educated spouse within ethnic group, the marriage rates among

foreign-born PhDs, especially for Chines and Indian PhDs in STEM occupations, are quite high. In Census 2000,

27

co-location problems may cause increasing concentration of married Ph.D. STEM workers in

large cities. This suggests that large cities may continue to be more thriving in innovative

activities versus smaller towns.25

Secondly, this paper implies that as ever-growing skilled immigrants arrive in the U.S.,

cities with a historically large ethnic population and those with denser educated ethnic

population are highly likely to experience a large-scale increase in human capital. That is

because such cities with appealing local marriage markets will attract the new-arrival highly

trained singles that are eager to find a life partner with similar education and ethnic backgrounds.

This increase of human capital might open great opportunities for such cities to develop

innovative activities. San Jose, CA, for example, in 1990 was a midsize MSA with a large size of

Chinese and Indian college-above population. Since then it has become one of the top destination

for Chinese and Indian Ph.D. workers in the U.S and experienced a tremendous amount of new

product innovation.26

Thirdly, this paper has provocative and controversial policy implications for declining

cities like Cleveland. To reinvent cities and achieve urban growth, policymakers in struggling

cities may find it appealing to invite highly trained immigrants by subsidizing Indian and

Chinese STEM workers with a tax credit for several years. At a certain point, these cities on their

own will start to attract new-arrival Ph.D.-trained individuals with similar ethnic backgrounds for

the active dating environment. The self-reinforcing growth of the highly educated ethnic

community may attract the firms that intensively use skilled workers to local areas, possibly

90% of Chinese Ph.D. STEM workers are married, 83% for Indian, 72% for the US-born and 64% for West

European. 25 This is consistent with the stylized fact documented by Kerr (2010): three large cities account for 25% of U.S.

domestic patenting for 1995-2004: San Francisco 12%, New York 7%, and Los Angeles 6%. 26 See more details in Appendix Table A6, which documents the MSA’s Chinese and Indian population in 1990

among the top 20 MSAs that create the largest number of patents in 2001.

28

generating agglomeration economies. However, this plan is also controversial because such

policies seem to be unfair to some immigrant groups like the Hispanic, who are more likely to

have the low education attainments.

7. Conclusions

Ph.D.-trained individuals, especially those that are not born in the U.S., increasingly

concentrate in large metropolitan areas from 1990 to 2010. This paper investigates the degree to

which the local marriage market and job colocation pressure affect the location choices of

foreign-born Ph.D. singles and couples differently compared to their US-born counterparts. This

work is important because Ph.D.-trained STEM workers account for a significant share of

innovation in the United States, and a sizable proportion of such Ph.D. workers are foreign born.

Moreover, foreign-born Ph.D. workers face with thin marriage markets compared to their U.S-

born counterparts. The contributing reason is that foreign-born PhDs, and especially those from

China and India, display a strong tendency to marry highly educated individuals within their own

ethnicity. This narrows the dating pool and summary measures confirm that there are relatively

few MSAs in the U.S. that offer active dating environment for these sorts of workers. For these

reasons, single, foreign-born Ph.D. workers may prefer to live in MSAs that can provide better

marriage markets.

Two key findings are as follows. First, marriage market opportunities attract immigrant

Ph.D.-trained singles to cities with denser potential dates. Chinese and Indian single Ph.D.

workers display a strong tendency to live in MSAs that have a large population of their own

ethnicity. The stronger desire to access an active dating environment explains Chinese and Indian

Ph.D. workers’ greater tendency relative to their US-born counterparts to live in midsize MSAs

29

with large sizes of own-ethnic college-above population. The marriage market effect is 4 to 7

percentage points for the female group, larger than that for the male group, roughly 2 percentage

points. This gender disparity is consistent with the asymmetry of marriage markets between

Ph.D.-trained men and women – it is more common for Ph.D.-trained men to marry less-

educated women than the reverse situation. Secondly, immigrant married Ph.D. workers that

have an own-ethnic Ph.D. spouse display a stronger tendency relative to their US-born

counterparts to live in large MSAs, regardless of the size of the local own-ethnic population.

That is mainly because large cities with better labor markets often enable both spouses in the

Super-Power Couples to find a job that is well matched to their training and within a reasonable

commute from their home. Such a job co-location effect seems to dominate the marriage market

effect in explaining foreign-born PhDs’ greater increasing trend to live in large cities.

30

Reference

Altonji, J.G., Card, D., 1991. The effects of immigration on the labor market outcomes of less-

skilled natives. In JM Abowd, RB Freeman, eds., Immigration, Trade, and the Labor Market, pp.

201–34. Chicago: Univ. Chicago Press.

Borjas, G., 2001. Does immigration grease the wheels of the labor market? Brookings Papers on

Economic Activity 2001(1), 69-119.

Burdett, K., Coles, M.G., 1999. Long-term partnership formation: marriage and employment.

Economic Journal 109, 307–334.

Card, D., 2001. Immigrant inflows, native outflows, and the local labor market impacts of higher

immigration. Journal of Labor Economics 19(1), 22-64.

Chen, Y., Rosenthal, S.S., 2008. Local amenities and life cycle migration: do people move for

jobs or fun? Journal of Urban Economics 65, 519–537.

Compton, J., Pollak, R., 2007. Why are power couples increasingly concentrated in large

metropolitan areas? Journal of Labor Economics 25, 475–512.

Costa, D.L., Kahn, M.E., 2000. Power couples: changes in the locational choice of the college

educated, 1940–1990. Quarterly Journal of Economics, 1287–1315.

Dahl, M.S., Sorenson, O., 2010. The migration of technical workers. Journal of Urban

Economics 67(1), 33–45.

Fang, D., Brown, D., 1999. Geographic mobility of the foreign-born Chinese in large

metropolises, 1985–1990. International Migration Review 33, 137–155.

Gautier, P.A., Svarer, M., Teulings, C.N., 2010. Marriage and the city: search frictions and

sorting of singles. Journal of Urban Economics 67(2), 206–218.

Gottlieb, P.D., Joseph, G., 2006. College-to-work migration of technology graduates and holders

of doctorates within the United States. Journal of Regional Science 46 (4), 627-659.

Hunt, J., Gauthier-Loiselle, M., 2010. How much does immigration boost innovation? American

Economic Journal: Macroeconomics 2, 31–56.

IPUMS-USA, University of Minnesota, www.ipums.org.

Kerr, W., 2010. Breakthrough inventions and migrating clusters of innovation. Journal of Urban

Economics 67(1), 46-60.

Kerr, W., Lincoln, W., 2010. The supply side of innovation: H-1B visas and US ethnic invention.

Journal of Labor Economics 28, 473–508.

31

Mandorff, M., 2007. Social networks, ethnicity, and occupation. University of Chicago. Ph.D.

Dissertation.

Peri, G., 2011. Rethinking the area approach: immigrants and the labor market in California.

Journal of International Economics 84(1), 1-14.

Steven Ruggles, J. Trent Alexander, Katie Genadek, Ronald Goeken, Matthew B. Schroeder, and

Matthew Sobek. Integrated Public Use Microdata Series: Version 5.0 [IPUMS-USA].

Minneapolis: University of Minnesota, 2010.

Walsh, J., Nagaoka, S., 2009. Who Invents?: Evidence from the Japan-U.S. inventor survey.

RIETI DP 09-E-034.

32

a New-Arrival refers to those foreign-born that have been in US less or equal to ten years. In this figure, San Jose, CA is excluded for visual clarity. In 2000, San Jose has 1,143

Chinese and 335 Indian New-Arrival Ph.D. workers in STEM Occupations. In 2001, 4,734 patents were created in San Jose, CA, and 276 are top 10% cited patents.

0

10

20

30

40

50

60

70

80

0

200

400

600

800

1000

1200

1400

1600

New-arrival Indian PhD STEM workers,Yr2000

New-arrival Chinese PhD STEM workers,Yr2000

Total patents,Yr2001

Top 10% cited Patents,Yr2001

Figure 1: New-Arrival Ph.D. STEM Workers at Top 30 MSAs in Terms of High-quality Patents Creation at Year 2001 N

um

ber o

f To

p 1

0%

Cited

Paten

ts Created

at 20

01

MS

A C

ou

nt

of W

ork

ers in 2

00

0 o

r of P

atents C

reatedin

20

01

33

Table 1: Marriage Patterns for Ph.D. Holders by Birth Region (BR), Census Year 2000 a

Panel A: Marriage Patterns for Male Ph.D. Holders, Age 25 to 65

Frequency of Wife’s Education Level and Birth Region (BR)

(Numbers Are in Percentage Points)

(1) (2) (3) (4) (5) (6)

Male PhDs’

Birth Region b

Same-BR

Wife

US-Born

Wife

Wife with

a Ph.D. Degree

Wife with Col+

Education

Same-BR Wife with

a Ph.D. Degree

Same-BR Wife with

Col+ Education

US-Born 92.8 92.8 12.3 77.1 10.8 71.5

Foreign-born 71.6 19.6 15.1 80.7 10.0 57.5

China 95.1 2.4 13.6 84.7 12.8 80.6

India 83.9 10.4 14.9 86.9 12.0 72.6

Korea 94.7 3.6 8.6 85.8 7.9 81.8

Philippines 71.8 21.4 16.9 82.2 14.1 64.9

Eastern Europe 78.6 13.8 22.1 86.7 17.5 70.0

Western Europe 41.2 42.3 16.1 81.7 6.7 31.9

Panel B: Marriage Patterns for Female Ph.D. Holders, Age 25 to 65

Female PhDs’

Birth Region b

Frequency of Husband’s Education and Birth Region (BR)

(Numbers Are in Percentage Points)

(1) (2) (3) (4) (5) (6)

Same-BR

Husband

US-Born

Husband

Husband with a

Ph.D. Degree

Husband with Col+

Education

Same-BR Husband

with a Ph.D. Degree

Same-BR Husband

with Col+ Education

US-Born 93.1 93.1 31.6 83.9 28.2 77.9

Foreign-born 59.7 29.9 48.0 89.8 30.2 53.6

China 78.5 15.2 56.2 96.3 44.1 75.8

India 83.6 12.2 51.6 96.3 44.9 81.1

Korea 66.2 24.7 54.8 90.1 33.7 60.5

Philippines 51.4 40.0 27.0 76.0 14.3 42.1

Eastern Europe 70.6 18.9 54.8 90.7 42.1 65.2

Western Europe 29.8 56.7 46.6 94.3 17.6 29.2 a Data are 5% sample decennial census 2000 extracted from IPUMS (Ruggles et al, 2010). This table is confined to the married Ph.D. holders who are living in married-couple

households in which both spouses are present and have reported their education levels, nativities and places of birth. b US-Born is defined as born in the U.S.A. or those born abroad with American Parent(s) and all the others are defined as foreign-born. I group countries into 22 regions mostly

based on United Nations Statistics Division. c Col+ (College-above) education refers to having received at least 4-year college education attainment.

34

Table 2-1: Top 25 MSAs with Largest Population of New-arrival Ph.D. Workers from China, Year 2010

Count of New-Arrival Chinese

Ph.D. Workers

Count of New-Arrival Chinese

Ph.D. Workers in STEM Occupations

Rank MSA Name Total Male Female Total

Married

Male

Married

Female

Single

Male

Single

Female

1 Washington, DC/MD/VA 3106 2092 1014 2798 1679 924 195 0

2 Boston, MA 2484 1362 1122 1833 832 658 343 0

3 Houston-Brazoria, TX 2088 1701 387 1201 864 262 75 0

4 New York-Northeastern NJ 1796 1347 449 1108 666 341 101 0

5 Chicago-Gary-Lake, IL 1774 934 840 1409 828 486 0 95

6 San Jose, CA 1751 1334 417 1413 1005 362 46 0

7 Philadelphia, PA/NJ 1427 273 1154 955 0 817 0 0

8 Oakland, CA 1395 945 450 1030 547 217 266 0

9 Baltimore, MD 1248 834 414 738 271 0 270 197

10 Atlanta, GA 1208 665 543 926 480 446 0 0

11 Raleigh-Durham, NC 1188 792 396 707 512 0 0 195

12 Los Angeles-Long Beach, CA 1170 677 493 888 477 142 200 69

13 Dallas-Fort Worth, TX 962 803 159 881 573 78 230 0

14 Seattle-Everett, WA 914 496 418 506 425 81 0 0

15 Newark, NJ 863 636 227 608 400 115 93 0

16 Detroit, MI 750 522 228 574 181 121 70 0

17 San Francisco-Oakland-

Vallejo, CA

718 291 427 718 194 354 97 73

18 Riverside-San Bernardino, CA 698 400 298 469 241 177 0 51

19 San Diego, CA 695 487 208 391 264 57 0 70

20 Miami-Hialeah, FL 623 484 139 464 308 83 73 0

21 Cleveland, OH 621 511 110 451 386 0 65 0

22 Ann Arbor, MI 519 350 169 430 261 169 0 0

23 Birmingham, AL 502 224 278 224 224 0 0 0

24 Austin, TX 497 392 105 260 199 0 0 0

25 Cincinnati OH/KY/IN 485 393 92 196 104 92 0 0 a Data are 1% sample from American Community Surveys 2010, extracted from IPUMS (Ruggles et al, 2010). New-arrival Ph.D. workers from China are immigrants who hold a

Ph.D. degree, born in China, aged 25 and above, employed at the survey year, and had been in the U.S. less or equal to ten years. Person weights are used to calculate these

numbers. b STEM Occupations are classified based on Census code list for STEM and STEM-related occupation, which is available at www.bls.gov/soc.

35

Table 2-2: Top 25 MSAs with Largest Population of New-arrival Ph.D. Workers from India, Year 2010

Count of New-Arrival Indian

Ph.D. Workers

Count of New-Arrival Indian

Ph.D. Workers in STEM Occupations

Rank MSA Name Total Male Female Total

Married

Male

Married

Female

Single

Male

Single

Female

1 Chicago-Gary-Lake, IL 1296 666 630 476 208 0 0 268

2 Boston, MA 1276 1055 221 896 666 0 230 0

3 Los Angeles-Long Beach, CA 1239 650 589 882 276 334 272 0

4 Philadelphia, PA/NJ 1207 965 242 769 457 80 135 97

5 Minneapolis-St. Paul, MN 976 582 394 133 0 133 0 0

6 State College, PA 845 845 0 845 0 0 845 0

7 San Diego, CA 777 530 247 530 429 0 101 0

8 Houston-Brazoria, TX 725 487 238 164 164 0 0 0

9 St. Louis, MO-IL 695 417 278 550 272 278 0 0

10 Oakland, CA 616 462 154 514 78 154 282 0

11 Dayton-Springfield, OH 603 603 0 0 0 0 0 0

12 Washington, DC/MD/VA 577 521 56 304 248 56 0 0

13 Atlanta, GA 567 420 147 500 270 147 83 0

14 New York-Northeastern NJ 543 305 238 344 0 116 106 122

15 San Jose, CA 531 468 63 383 320 63 0 0

16 Portland-Vancouver, OR 466 466 0 466 466 0 0 0

17 Trenton, NJ 450 236 214 249 141 108 0 0

18 Baltimore, MD 435 61 374 61 61 0 0 0

19 Cleveland, OH 430 381 49 430 166 49 215 0

20 Austin, TX 411 316 95 54 54 0 0 0

21 Charlottesville, VA 405 0 405 405 0 135 0 270

22 Seattle-Everett, WA 390 309 81 321 132 81 108 0

23 Mobile, AL 364 364 0 364 364 0 0 0

24 Lowell, MA/NH 348 348 0 0 0 0 0 0

25 Dallas-Fort Worth, TX 339 339 0 196 196 0 0 0 a Data are 1% sample from American Community Surveys 2010, extracted from IPUMS (Ruggles et al, 2010). New-arrival Ph.D. workers from India are immigrants who hold a

Ph.D. degree, born in India, aged 25 and above, employed at the survey year, and had been in U.S. less or equal to ten years. Person weights are used to calculate these numbers. b STEM Occupations are classified based on Census code list for STEM and STEM-related occupation, which is available at www.bls.gov/soc.

36

Table 3-1: Location Choices for Single Ph.D. Workers by Year,

Classify City Sizes Based on MSA’s Total Population

City Size Categories Based

on MSA’s Total Population Single Ph.D. Workers (SPS)

Predicted Probabilities to Live in Different City Sizes

(1) (2) (3)

Year 1990 Year 2000 Year 2010

Panel A: Single, US-born White Male PhDs Large MSAs 0.2644 (.00014) 0.3316 (.00012) 0.4126 (.00035)

Midsize MSAs 0.4584 (.00044) 0.4289 (.00044) 0.3714 (.00084)

Small MSAs 0.2772 (.00037) 0.2395 (.00035) 0.2160 (.00069)

Panel B: Single, US-born White Female PhDs Large MSAs 0.2678 (.00026) 0.3480 (.00017) 0.3721 (.00031)

Midsize MSAs 0.4355 (.00048) 0.4117 (.00028) 0.4080 (.00050)

Small MSAs 0.2967 (.00064) 0.2403 (.00029) 0.2199 (.00041)

Panel C: Single, Chinese Male PhDs Large MSAs 0.3238 (.00400) 0.3917 (.00277) 0.4977 (.00598)

Midsize MSAs 0.4673 (.00408) 0.4445 (.00259) 0.3783 (.00424)

Small MSAs 0.2089 (.00609) 0.1638 (.00363) 0.1240 (.00324)

Panel D: Single, Chinese Female PhDs Large MSAs 0.3751 (.00796) 0.4326 (.00317) 0.4693 (.01038)

Midsize MSAs 0.4490 (.00893) 0.4399 (.00207) 0.3411 (.00763)

Small MSAs 0.1759 (.00834) 0.1275 (.00419) 0.1896 (.00868)

Panel E: Single, Indian Male PhDs Large MSAs 0.3106 (.00322) 0.3991 (.00228) 0.4365 (.00564)

Midsize MSAs 0.4359 (.00360) 0.4628 (.00248) 0.4489 (.00411)

Small MSAs 0.2535 (.00577) 0.1381 (.00264) 0.1146 (.00272)

Panel F: Single, Indian Female PhDs Large MSAs 0.3376 (.00831) 0.4020 (.00400) 0.4365 (.00778)

Midsize MSAs 0.4123 (.01133) 0.4333 (.00258) 0.3568 (.00499)

Small MSAs 0.2501 (.01313) 0.1647 (.00591) 0.2167 (.01140) a City size categories are based on the size of MSA’s total population for those who are aged 15 above. Large MSAs are

Metropolitan areas in which the total population for those who are aged 15 above was at least 2 million, midsize MSAs as those

between 2 million and 250,000, and small MSAs as those less than 250,000 or in rural areas. Standard errors are in parentheses. b Data for 1990 and 2000 are 5% sample from decennial census, and for 2010 are 1% sample from ACS. For each year, the

probabilities should sum to one within panel. Standard errors are in parentheses.

37

Table 3-2: Location Choices for Married Ph.D. Workers in Super-Power Couples (SPC)

Classify City Sizes Based on MSA’s Total Population

City Size Categories Based

on MSA’s Total Population Married Ph.D. Workers (SPC)

Predicted Probabilities to Live in Different City Sizes

(1) (2) (3)

Year 1990 Year 2000 Year 2010

Panel A: Married, US-born White Male PhDs Large MSAs 0.2729 (.00024) 0.3011 (.00037) 0.3329 (.00093)

Midsize MSAs 0.4241 (.00067) 0.4321 (.00058) 0.3944 (.00103)

Small MSAs 0.3030 (.00075) 0.2668 (.00054) 0.2727 (.00090)

Panel B: Married, US-born White Female PhDs Large MSAs 0.2575 (.00056) 0.3010 (.00031) 0.3331 (.00065)

Midsize MSAs 0.4389 (.00082) 0.4395 (.00069) 0.3895 (.00136)

Small MSAs 0.3036 (.00130) 0.2595 (.00082) 0.2774 (.00154)

Panel C: Married, Chinese Male PhDs Large MSAs 0.3029 (.00499) 0.3703 (.00183) 0.4937 (.00273)

Midsize MSAs 0.5229 (.00407) 0.4957 (.00105) 0.3392 (.00141)

Small MSAs 0.1742 (.00656) 0.1340 (.00183) 0.1671 (.00186)

Panel D: Married, Chinese Female PhDs Large MSAs 0.2649 (.00505) 0.3737 (.00166) 0.4893 (.00327)

Midsize MSAs 0.5878 (.00592) 0.5076 (.00168) 0.3482 (.00252)

Small MSAs 0.1473 (.00547) 0.1187 (.00170) 0.1625 (.00244)

Panel E: Married, Indian Male PhDs Large MSAs 0.2897 (.00375) 0.3312 (.00239) 0.4507 (.00573)

Midsize MSAs 0.4981 (.00319) 0.5206 (.00174) 0.3881 (.00310)

Small MSAs 0.2122 (.00499) 0.1482 (.00274) 0.1612 (.00342)

Panel F: Married, Indian Female PhDs Large MSAs 0.2716 (.00283) 0.3326 (.00246) 0.4242 (.00406)

Midsize MSAs 0.4940 (.00486) 0.5022 (.00228) 0.3815 (.00412)

Small MSAs 0.2344 (.00622) 0.1652 (.00278) 0.1943 (.00378) a City size categories are based on the size of MSA’s total population for those who are aged 15 above. Large MSAs are

Metropolitan areas in which the total population for those who are aged 15 above was at least 2 million, midsize MSAs as those

between 2 million and 250,000, and small MSAs as those less than 250,000 or in rural areas. For each year, the probabilities

should sum to one within panel. Standard errors are in parentheses. b Data for 1990 and 2000 are 5% sample from decennial census, and for 2010 are 1% sample from ACS. Predicted Probabilities

are averaged among married Ph.D. workers that have a Ph.D.-holding spouse within their own ethnicity.

38

Table 4: Differencing Strategy to Identify the Marriage Market and Job Co-Location Effects:

Using Chinese as an Example

Household Type a Label Benefits of living in a large city b

1 Chinese Super-Power Couples

(SPC)

SPC𝐶ℎ Job Co-Location 𝐶𝐶ℎ,

Urban Amenities 𝐴𝐶ℎ,

Returns to Education

2 US-born Super-Power Couples

SPC𝑈𝑆 Job Co-Location 𝐶𝑈𝑆,

Urban Amenities 𝐴𝑈𝑆,

Returns to Education

3 Double Difference: Row1- Row2

SPC𝐶ℎ − SPC𝑈𝑆 Urban Amenities (𝐴𝐶ℎ − 𝐴𝑈𝑆 ) +

Job Co-Location (𝐶𝐶ℎ − 𝐶𝑈𝑆)

4 Chinese Super-Power Singles (SPS)

SPS𝐶ℎ Marriage Markets 𝑀𝐶ℎ,

Urban Amenities 𝐴𝐶ℎ,

Returns to Education

5 US-born Super-Power Singles

SPS𝑈𝑆 Marriage Markets 𝑀𝑈𝑆,

Urban Amenities 𝐴𝑈𝑆,

Returns to Education

6 Double Difference: Row4 - Row5

SPS𝐶ℎ − SPS𝑈𝑆 Urban amenities (𝐴𝐶ℎ − 𝐴𝑈𝑆) +

Marriage Markets (𝑀𝐶ℎ −𝑀𝑈𝑆)

7 Triple Difference: Row6 - Row3

(SPS𝐶ℎ − SPS𝑈𝑆)− (SPC𝐶ℎ − SPC𝑈𝑆)

Marriage Markets (𝑀𝐶ℎ −𝑀𝑈𝑆)–

Job Co-Location (𝐶𝐶ℎ − 𝐶𝑈𝑆)

a “Super-Power” Couples: married couple households where both spouses in the family have a Ph.D. degree. “Super-Power”

Singles (SPS): Ph.D. workers who report their marital status as never married, divorced or widowed. b Within the gender group, returns to education for Ph.D. holders in a given city type are assumed same between Chinese and US

born, regardless of the marital status. Chinese and US-born Ph.D. workers may have different tastes for urban amenities (like

access to cultural activities), but the difference between their valuations on urban amenity (𝐴𝑐ℎ − 𝐴𝑈𝑆) are assumed to stay the

same across household types.

39

Table 5-1: Trends in Propensity to Live in Different City Sizes, Year 1990 to 2000,

Classify City Sizes Based on MSA’s Total Population

City Size Category Based on

MSA’s Total Population

Female Ph.D. Workers Male Ph.D. Workers

(1) (2) (3) (4)

Single Ph.D. Workers

(SPS)

Married Ph.D. Workers

in SPC

Single Ph.D. Workers

(SPS)

Married Ph.D. Workers

in SPC

Panel A: US-born

SPS𝑈𝑆 SPC𝑈𝑆 SPS𝑈𝑆 SPC𝑈𝑆

Large MSAs 0.0802*** 0.0435*** 0.0672*** 0.0282***

(.00031) (.00064) (.00018) (.00044)

Midsize MSAs -0.0239*** 0.0006 -0.0295*** 0.0080***

(.00055) (.00107) (.00063) (.00088)

Small MSAs -0.0563*** -0.0441*** -0.0377*** -0.0362***

(.00070) (.00154) (.00051) (.00092)

Panel B: Chinese

𝑆𝑃𝑆𝐶ℎ SPC𝐶ℎ 𝑆𝑃𝑆𝐶ℎ SPC𝐶ℎ

Large MSAs 0.0575*** 0.1088*** 0.0679*** 0.0674***

(.00856) (.00457) (.00487) (.00539)

Midsize MSAs -0.0092 -0.0802*** -0.0228*** -0.0272***

(.00917) (.00616) (.00480) (.00421)

Small MSAs -0.0483*** -0.0286*** -0.0451*** -0.0402***

(.00899) (.00573) (.00686) (.00560)

Panel C: Indian

𝑆𝑃𝑆𝐼𝑛 SPC𝐼𝑛 𝑆𝑃𝑆𝐼𝑛 SPC𝐼𝑛

Large MSAs 0.0644*** 0.0610*** 0.0885*** 0.0415***

(.00933) (.00375) (.00386) (.00427)

Midsize MSAs 0.0210* 0.0082 0.0269*** 0.0225***

(.01162) (.00537) (.00425) (.00364)

Small MSAs -0.0854*** -0.0692*** -0.1154*** -0.0640***

(.01395) (.00681) (.00634) (.00569) a City size categories are based on the size of MSA’s total population for those who are aged 15 above. Large MSAs are metropolitan areas in which the total population for those

who are aged 15 above was at least 2 million, midsize MSAs as those between 2 million and 250,000, and small MSAs as those less than 250,000 or in rural areas. Standard errors

are in parentheses. ***denotes the estimate is significant from zero at 1% level, ** at 5% level and * at 10% level. b The first difference estimators for Single Ph.D. workers (SPS) are obtained by subtracting Column 1 from Column 2 in Table 3-1. Similarly, subtracting Column 1 from Column

2 in Table 3-2 obtained the first-differencing estimators for Super-Power Couples.

40

Table 5-2: DOUBLE DIFFERENCE Estimators by Gender, Year 1990 to 2000,

Classify City Sizes Based on MSA’s Total Population

Panel A: Differential Trends Between Chinese and US-born

City Size Category Based on

MSA’s Total Population

Female Ph.D. Workers Male Ph.D. Workers

(1) (2) (3) (4)

Single PhDs Married PhDs Single PhDs Married PhDs

(SPS𝐶ℎ − SPS𝑈𝑆) (SPC𝐶ℎ − SPC𝑈𝑆) (SPS𝐶ℎ − SPS𝑈𝑆) (SPC𝐶ℎ − SPC𝑈𝑆) Large MSAs -0.0227*** 0.0653*** 0.0007 0.0392***

(.00726) (.00305) (.00419) (.00225)

Midsize MSAs 0.0147*** -0.0808*** 0.0067*** -0.0352***

(.00439) (.00628) (.00217) (.00414)

Small MSAs 0.0080* 0.0155*** -0.0074** -0.0040**

(.00426) (.00450) (.00348) (.00202)

Panel B: Differential Trends Between Indian and US-born

City Size Category Based on

MSA’s Total Population

Female Ph.D. Workers Male Ph.D. Workers

(1) (2) (3) (4)

Single PhDs Married PhDs Single PhDs Married PhDs

(𝑆𝑃𝑆𝐼𝑛 − 𝑆𝑃𝑆𝑈𝑆) (𝑆𝑃𝐶𝐼𝑛 − 𝑆𝑃𝐶𝑈𝑆) (𝑆𝑃𝑆𝐼𝑛 − 𝑆𝑃𝑆𝑈𝑆) (𝑆𝑃𝐶𝐼𝑛 − 𝑆𝑃𝐶𝑈𝑆) Large MSAs -0.0158*** 0.0175*** 0.0213*** 0.0133***

(.00469) (.00353) (.00353) (.00260)

Midsize MSAs 0.0449*** 0.0076** 0.0564*** 0.0145***

(.00599) (.00351) (.00210) (.00244)

Small MSAs -0.0291*** -0.0251*** -0.0777*** -0.0278***

(.00661) (.00579) (.00650) (.00470) a City size categories are based on MSA’s total population for those who are aged 15 above. Large MSAs are Metropolitan areas in which the total population for those who are

aged 15 above was at least 2 million, midsize MSAs as those between 2 million and 250,000, and small MSAs as those less than 250,000 or in rural areas. Standard errors are in

parentheses. b The double difference estimators are obtained by differencing trends in propensity to Live in a given city size between foreign and domestic born Ph.D. workers, conditional on

gender and household type. Standard errors are in parentheses. And ***denotes the estimate is significant from zero at 1% level, ** at 5% level and * at 10% level. For each panel

and within each column, the probabilities should sum to zero.

41

Table 5-3: TRIPLE DIFFERENCE Estimators by Gender

City Sizes Based on MSA’s Total Population

City Size Category Based on

MSA’s Total Population

Triple Difference,

Marriage Markets Effects Relative to Colocation Effects

Year 1990 to 2000

Year 1990 to 2010

(1) (2) (3) (4)

Female Ph.D. Workers Male Ph.D. Workers Female Ph.D. Workers Male Ph.D. Workers

Panel A: Chinese Relative to US-born

(SPS𝐶ℎ − SPS𝑈𝑆) − (SPC𝐶ℎ − SPC𝑈𝑆)

Large MSAs -0.0881*** -0.0385*** -0.1588*** -0.1052***

(.00234) (.00355) (.00841) (.00859)

Midsize MSAs 0.0955*** 0.0419*** 0.1098*** 0.1520***

(.00498) (.00058) (.00837) (.00866)

Small MSAs -0.0075 -0.0034 0.0490*** -0.0468

(.00735) (.00217) (.00271) (.00373)

Panel B: Indian Relative to US-born

(𝑆𝑃𝑆𝐼𝑛 − 𝑆𝑃𝑆𝑈𝑆) − (𝑆𝑃𝐶𝐼𝑛 − 𝑆𝑃𝐶𝑈𝑆)

Large MSAs -0.0333*** 0.0079 -0.0924*** -0.1232***

(.00268) (.00285) (.00560) (.00793)

Midsize MSAs 0.0373*** 0.0419*** 0.0351*** 0.1801***

(.00065) (.00058) (.00771) (.00590)

Small MSAs -0.0040 -0.0498*** 0.0573*** -0.0569***

(.00439) (.00434) (.00772) (.00519) a City size categories are based on the MSA’s total population for those who are aged 15 above. Large MSAs are Metropolitan areas in which the total population for those who

are aged 15 above is at least 2 million, midsize MSAs as those between 2 million and 250,000, and small MSAs as those less than 250,000 or in rural areas. b Each cell is obtained by differencing the double-differencing estimators across household type, within gender group. Standard errors are in parentheses. ***denotes the estimate

is significant from zero at 1% level, ** at 5% level and * at 10% level. Within each panel and for a given column, the probabilities should sum to zero.

42

Table 6: City Size Categories Based on MSA’s Own-Ethnic Population,

TRIPLE DIFFERENCE Estimators by Gender

City Size Category Based on

MSA’s Own-Ethnic Population

Marriage Markets Effects Relative to Colocation Effects

Year 1990 to 2000

Year 1990 to 2010

(1) (2) (3) (4)

Female Ph.D. Workers Male Ph.D. Workers Female Ph.D. Workers Male Ph.D. Workers

Panel A: Chinese Relative to US-born

(SPS𝐶ℎ − SPS𝑈𝑆) −(SPC𝐶ℎ − SPC𝑈𝑆)

Large Own-Ethnic MSAs 0.0014 -0.0606*** -0.1090*** -0.0844***

(.00289) (.00227) (.00419) (.00367)

Midsize Own-Ethnic MSAs -0.0369 0.0336*** 0.0311*** 0.0802***

(.00522) (.00249) (.00822) (.00391)

Small Own-Ethnic MSAs 0.0355*** 0.0270*** 0.0779*** 0.0042

(.00104) (.00552) (.00647) (.00733)

Panel B: Indian Relative to US-born

(𝑆𝑃𝑆𝐼𝑛 − 𝑆𝑃𝑆𝑈𝑆) − (𝑆𝑃𝐶𝐼𝑛 − 𝑆𝑃𝐶𝑈𝑆)

Large Own-Ethnic MSAs 0.0279*** -0.0261*** -0.0521*** -0.1504***

(.00421) (.00317) (.00590) (.00573)

Midsize Own-Ethnic MSAs -0.1502*** 0.0305*** -0.1724*** 0.1442***

(.00420) (.00361) (.01313) (.00827)

Small Own-Ethnic MSAs 0.1223*** -0.0044 0.2245*** 0.0062

(.00854) (.00555) (.00903) (.00879) a City size categories are based on the size of the own-ethnic population that are aged 15 above and present in that MSA. For a given person, large own-ethnic MSAs are

metropolitan areas in which his/her own ethnic population rank within top 15 among all U.S. MSAs, midsize own-ethnic MSAs as those that rank outside top 15 but have at least

1,000 people within the ethnicity of this person (the cutoff number is 250,000 for the US-born) present in the local areas, and the rest of MSAs are classified as small own-ethnic

MSAs. b Within a specific gender-period, for each panel, the probabilities should sum to zero. Standard errors are in parentheses. ***denotes the estimate is significant from zero at 1%

level, ** at 5% level and * at 10% level.

43

Table 7-1: City Size Categories Based on

MSA’s Total Population and Own-Ethnic Col+ Population (OECoAbv),

Trends of Location Choices for Single Ph.D. Workers

City Size Category Based on

MSA’s Total Population and Own-

Ethnic Col+ Population (OECoAbv)

Year 1990 to 2000

Year 1990 to 2010

(1) (2) (3) (4)

Single Female PhDs Single Male PhDs Single Female PhDs Single Male PhDs

Panel A: US-born (SPS𝑈𝑆)

Large MSAs + Large OECoAbv 0.0385*** (.00286) 0.0243*** (.00020) 0.0268*** (.00043) 0.0668*** (.00033)

Large MSAs + Small OECoAbv 0.0416*** (.00008) 0.0426*** (.00008) 0.0779*** (.00011) 0.0812*** (.00027)

Midsize MSAs + Large OECoAbv -0.0343*** (.00013) -0.0241*** (.00011) -0.0632*** (.00012) -0.0506*** (.00007)

Midsize MSAs + Small OECoAbv 0.0105*** (.00045) -0.0051*** (.00052) 0.0353*** (.00063) -0.0361*** (.00094)

Small MSAs -0.0563*** (.00070) -0.0377*** (.00051) -0.0768*** (.00076) -0.0613*** (.00078)

Panel B: Chinese (SPS𝐶ℎ)

Large MSAs + Large OECoAbv 0.0095 (.00784) 0.0207*** (.00463) 0.0001 (.01314) 0.1038*** (.00870)

Large MSAs + Small OECoAbv 0.0454*** (.00121) 0.0483*** (.00113) 0.0830*** (.00343) 0.0712*** (.00373)

Midsize MSAs + Large OECoAbv 0.0226*** (.00696) 0.0153*** (.00443) -0.1019*** (.00764) -0.0838*** (.00451)

Midsize MSAs + Small OECoAbv -0.0273*** (.00444) -0.0387*** (.00239) 0.0087*** (.00745) -0.0059 (.00589)

Small MSAs -0.0502*** (.00920) -0.0456*** (.00696) 0.0101 (.01240) -0.0853*** (.00689)

Panel C: Indian (SPS𝐼𝑛)

Large MSAs + Large OECoAbv 0.0749*** (.01063) 0.0752*** (.00371) 0.0588*** (.00848) 0.0450*** (.00843)

Large MSAs + Small OECoAbv -0.0016 (.00360) 0.0121*** (.00141) 0.0404*** (.00556) 0.0824*** (.00422)

Midsize MSAs + Large OECoAbv 0.0278*** (.00385) 0.0078*** (.00272) 0.0004 (.00598) -0.0524*** (.00346)

Midsize MSAs + Small OECoAbv -0.0202** (.00773) 0.0186*** (.00175) -0.0717*** (.00783) 0.0629*** (.00590)

Small MSAs -0.0809*** (.01349) -0.1137*** (.00624) -0.0279 (.01709) -0.1379*** (.00616) a City size categories are based on two indicators :1) the MSA’s total population who aged 15 above, and on 2) the size of the own-ethnic population that have at least four-year

college education levels (OECoAbv) and are present in that MSA. Large MSAs are metropolitan areas in which the total population for those who are aged 15 above was at least 2

million, midsize MSAs as those between 2 million and 250,000, and small MSAs as those less than 250,000 or in rural areas. For a Ph.D. worker, an MSA is defined as having

Large OECoAbv if the size of the own-ethnic college-above population in that MSA rank within top 15 among all U.S. MSAs in a given survey year; otherwise, an MSA is

defined as having Small OECoAbv. b For a gender group in a specific period, the probabilities should sum to zero within each panel. Standard errors are in parentheses. ***denotes the estimate is significant from zero

at 1% level, ** at 5% level and * at 10% level.

44

Table 7-2: City Size Categories Based on

MSA’s Total Population and Own-Ethnic Col+ Population (OECoAbv),

Trends of Location Choices for Married Ph.D. Workers in Super-Power Couples (SPC)

City Size Category Based on

MSA’s Total Population and Own-

Ethnic Col+ Population (OECoAbv)

Year 1990 to 2000

Year 1990 to 2010

(1) (2) (3) (4) Married Female PhDs

in SPC

Married Male PhDs

in SPC

Married Female PhDs

in SPC

Married Male PhDs

in SPC

Panel A: US-born (SPC𝑈𝑆)

Large MSAs + Large OECoAbv 0.0131*** (.00062) -0.0001 (.00041) 0.0236*** (.00084) 0.0062*** (.00091)

Large MSAs + Small OECoAbv 0.0301*** (.00007) 0.0288*** (.00008) 0.0526*** (.00019) 0.0537*** (.00027)

Midsize MSAs + Large OECoAbv -0.0208*** (.00026) -0.0186*** (.00021) -0.0565*** (.00015) -0.0517*** (.00013)

Midsize MSAs + Small OECoAbv 0.0220*** (.00085) 0.0262*** (.00072) 0.0070*** (.00015) 0.0220*** (.00120)

Small MSAs -0.0444*** (.00154) -0.0363*** (.00092) -0.0267*** (.00202) -0.0302*** (.00117)

Panel B: Chinese (SPC𝐶ℎ)

Large MSAs + Large OECoAbv 0.0566*** (.00341) 0.0181*** (.00562) 0.1097*** (.00509) 0.0760*** (.00520)

Large MSAs + Small OECoAbv 0.0487*** (.00207) 0.0446*** (.00115) 0.1118*** (.00355) 0.1075*** (.00242)

Midsize MSAs + Large OECoAbv -0.0322*** (.00666) -0.0009 (.00574) -0.1454*** (.00658) -0.1183*** (.00564)

Midsize MSAs + Small OECoAbv -0.0415*** (.00459) -0.0196*** (.00629) -0.0870*** (.00582) -0.0541*** (.00642)

Small MSAs -0.0316*** (.00588) -0.0422*** (.00564) 0.0109*** (.00533) -0.0111 (.00671)

Panel C: Indian (SPC𝐼𝑛)

Large MSAs + Large OECoAbv 0.0395*** (.00317) 0.0296*** (.00487) 0.0934*** (.00526) 0.0845*** (.00690)

Large MSAs + Small OECoAbv 0.0187*** (.00163) 0.0165*** (.00067) 0.0641*** (.00277) 0.0791*** (.00322)

Midsize MSAs + Large OECoAbv 0.0003 (.00274) 0.0224*** (.00326) -0.0467*** (.00298) -0.0462*** (.00318)

Midsize MSAs + Small OECoAbv 0.0105*** (.00336) -0.0062 (.00474) -0.0742*** (.00580) -0.0692*** (.00570)

Small MSAs -0.0690*** (.00657) -0.0623*** (.00485) -0.0366*** (.00708) -0.0482*** (.00548) a City size categories are based on two indicators :1) the MSA’s total population who aged 15 above, and on 2) the size of the own-ethnic population that have at least four-year

college education levels (OECoAbv) and are present in that MSA. Large MSAs are metropolitan areas in which the total population for those who are aged 15 above was at least 2

million, midsize MSAs as those between 2 million and 250,000, and small MSAs as those less than 250,000 or in rural areas. For a Ph.D. worker, an MSA is defined as having

Large OECoAbv if the size of the own-ethnic college-above population in that MSA rank within top 15 among all U.S. MSAs in a given survey year; otherwise, an MSA is

defined as having Small OECoAbv. b For a gender group in a specific period, the probabilities should sum to zero within each panel. Standard errors are in parentheses. ***denotes the estimate is significant from zero

at 1% level, ** at 5% level and * at 10% level.

45

Table 7-3: City Size Categories Based on

MSA’s Total Population and Own-Ethnic Col+ Population (OECoAbv),

TRIPLE DIFFERENCE Estimators by Gender

City Size Category Based on

MSA’s Total Population and Own-

Ethnic Col+ Population (OECoAbv)

Marriage Markets Effects

Year 1990 to 2000 Year 1990 to 2010 (1)

Female Ph.D. Workers

(2)

Male Ph.D. Workers

(3)

Female Ph.D. Workers

(4)

Male Ph.D. Workers

Panel A: Chinese Relative to US-born

(SPS𝐶ℎ − SPS𝑈𝑆) −(SPC𝐶ℎ − SPC𝑈𝑆)

Large MSAs + Large OECoAbv -0.0725*** (.00256) -0.0219*** (.00141) -0.1128*** (.00450) -0.0328*** (.00448)

Large MSAs + Small OECoAbv -0.0149*** (.00112) -0.0101*** (.00155) -0.0541*** (.00741) -0.0637*** (.00518)

Midsize MSAs + Large OECoAbv 0.0684*** (.00293) 0.0217*** (.00235) 0.0502*** (.00691) 0.0334*** (.00486)

Midsize MSAs + Small OECoAbv 0.0257*** (.00192) 0.0123*** (.00053) 0.0673*** (.00434) 0.1063*** (.00482)

Small MSAs -0.0067 (.00736) -0.0020 (.00218) 0.0494*** (.00264) -0.0432*** (.00373)

Panel B: Indian Relative to US-born

(𝑆𝑃𝑆𝐼𝑛 − 𝑆𝑃𝑆𝑈𝑆) − (𝑆𝑃𝐶𝐼𝑛 − 𝑆𝑃𝐶𝑈𝑆)

Large MSAs + Large OECoAbv 0.0100*** (.00296) 0.0211*** (.00174) -0.0378*** (.00505) -0.1001*** (.00573)

Large MSAs + Small OECoAbv -0.0319*** (.00473) -0.0182*** (.00232) -0.0490*** (.00290) -0.0240*** (.00668)

Midsize MSAs + Large OECoAbv 0.0411*** (.00242) -0.0091*** (.00185) 0.0538*** (.00384) -0.0074** (.00325)

Midsize MSAs + Small OECoAbv -0.0192*** (.00188) 0.0561*** (.00155) -0.0258*** (.00841) 0.1902*** (.00583)

Small MSAs 4.87e-05 (.00435) -0.0499*** (.00378) 0.0588*** (.00163) -0.0587*** (.00516) a City size categories are based on two indicators :1) the MSA’s total population who aged 15 above, and on 2) the size of the own-ethnic population that have at least four-year

college education levels (OECoAbv) and are present in that MSA. Large MSAs are metropolitan areas in which the total population for those who are aged 15 above was at least 2

million, midsize MSAs as those between 2 million and 250,000, and small MSAs as those less than 250,000 or in rural areas. For a Ph.D. worker, an MSA is defined as having

Large OECoAbv if the size of the own-ethnic college-above population in that MSA rank within top 15 among all U.S. MSAs in a given survey year; otherwise, an MSA is

defined as having Small OECoAbv. b For a gender group in a specific period, the probabilities should sum to zero within each panel. Standard errors are in parentheses. ***denotes the estimate is significant from zero

at 1% level, ** at 5% level and * at 10% level.

46

Table 8: Influence of Marriage and Children

Migration for Young (Age 25 to 44) Ph.D. Workers

Dependent Variable:

1 if Moved to a Different MSA, 0 if Stayed in the Same MSA.

All Single Ph.D. Workers +

Married Ph.D. Workers Who Made the Moving Decision Post-Marriage

With or Without Children Present at Home No Child Present at Home

(1) (2) (3) (4)

Female Male Female Male

Count of Own-Ethnic Col+ Single of the Opposite

Gender in 1,000s in the Previous MSA (M)

0.002 4.0e-04 0.001 4.2e-04

(0.002) (0.001) (0.002) (0.001)

Single -0.284 1.388 0.771 0.115

(1.497) (1.192) (1.809) (2.735)

Foreign-Born (=1 if the worker is not a US-Born) 0.013 0.021 0.015 0.014

(0.066) (0.039) (0.073) (0.049)

M × Single -9.3e-04 -4.2e-04 -6.1e-04 -5.4e-04

(7.4e-04) (5.6e-04) (9.2e-04) (8.2e-04)

M × Single × Foreign-Born -0.033* -0.033*** -0.038* -0.034***

(0.018) (0.009) (0.021) (0.009)

Has Own Child in the Household -0.072** -0.081** - -

(0.032) (0.038) - -

MSA Fixed Effects 236 252 228 247

Industry Fixed Effects 21 21 21 21

Year Fixed Effects Yes Yes Yes Yes

Individual Characteristics Yes Yes Yes Yes

R-squared 0.240 0.193 0.249 0.192

Observations 2,508 3,494 2,141 3,149 a Sample is restricted to fulltime Ph.D. workers whose usual hours worked per week ≥ 35hours. I suppress the coefficient estimates on the characteristics of the previous MSA

locations, e.g., total population, own-industry employment, number of own-ethnic workers excluding col+ single of the opposite gender, MSA average working age, MSA average

annual earnings. I also suppress the coefficient estimates on individual attributes, e.g. age, square of age, occupation's median earned income, Hispanic origin, Black race, English

Speaking Ability, years in the U.S.no more than ten, living in birth state, and also those on characteristics of the worker’s spouse, e.g. spouse’s education level, age, English

speaking ability. Standard errors are clustered at MSA level and are reported in parentheses. ***denotes the estimate is significant from zero at 1% level, ** at 5% level.

47

Appendix Table A1: Share of Workers Living in Large MSAs by Education Attainment, 1990-2010

Panel A: Workers Holding a Ph.D. Degree

Share of Workers Living in Large MSAs

(1)

Year 1990

(2)

Year 2000

(3)

Year 2010

All Ph.D. Workers 0.252 0.325 0.394

Female Ph.D. Workers:

All Female PhDs 0.285 0.352 0.407

US-born Female PhDs 0.267 0.329 0.378

Foreign-born Female PhDs 0.380 0.441 0.496

Male Ph.D. Workers:

All Male PhDs 0.242 0.313 0.385

US-born Male PhDs 0.220 0.281 0.338

Foreign-born Male PhDs 0.329 0.401 0.492

Panel B: Workers Having a College or Master Degree

Share of Workers Living in Large MSAs

(1)

Year 1990

(2)

Year 2000

(3)

Year 2010

All Workers 0.234 0.318 0.371

Female Workers:

All Female 0.239 0.316 0.368

US-born Female 0.223 0.294 0.341

Foreign-born Female 0.423 0.494 0.535

Male Workers:

All Male 0.230 0.320 0.375

US-born Male 0.212 0.297 0.347

Foreign-born Male 0.404 0.483 0.531

Panel C: Workers with High School or a Lower Level of Education

Share of Workers Living in Large MSAs

(1)

Year 1990

(2)

Year 2000

(3)

Year 2010

All Workers 0.166 0.232 0.282

Female Workers:

All Female 0.168 0.231 0.280

US-born Female 0.143 0.198 0.235

Foreign-born Female 0.413 0.479 0.496

Male Workers:

All Male 0.164 0.233 0.284

US-born Male 0.132 0.188 0.224

Foreign-born Male 0.433 0.491 0.498

48

Table A2-1: Distribution of Col+ Single Men among U.S. MSAs by Ethnicity (Birth Region):

Number of MSAs in 6 Size Categories a,

Size Categories Based on MSA Count of Own-Ethnic Col+ Single Men (M)

Year 1990 and Year 2010

(1) (2) (3) (4) (5) (6)

MSA Size Categories Are

Based on MSA Count of Own-

Ethnic Col+ Single Men (M)

200 ≤ M ≤ 499

500 ≤ M ≤ 999

1000 ≤ M ≤ 1999

2000 ≤ M ≤ 3999

4000 ≤ M ≤ 7999

M ≥ 8000

Year

1990

Year

2010

Year

1990

Year

2010

Year

1990

Year

2010

Year

1990

Year

2010

Year

1990

Year

2010

Year

1990

Year

2010

Ethnicity (Birth Region)

United States 2b 3 18 10 53 34 73 54 47 80 80 116

Canada 10 17 5 15 2 2 0 2 0 0 0 0

Other North America 0 0 0 0 0 0 0 0 0 0 0 0

Central America and Caribbean 16 31 11 16 5 14 0 14 2 4 1 3

Southern America 14 31 4 10 1 8 2 4 1 2 0 1

Northern Europe 16 18 6 14 3 6 1 1 0 1 0 0

Western Europe 13 26 8 10 2 3 1 1 0 1 0 0

Southern Europe 10 21 4 9 1 3 1 0 0 1 0 0

Eastern Europe 15 29 3 10 1 9 2 2 0 1 0 1

China 10 31 6 17 4 5 0 4 2 3 0 2

Japan 7 10 0 2 2 2 0 0 0 0 0 0

Korea 6 19 0 6 1 6 1 0 0 2 0 0

Other East Asia 0 2 0 0 0 0 0 0 0 0 0 0

Philippines 10 20 4 7 4 8 0 6 1 0 0 1

Vietnam 7 12 1 13 3 5 0 3 0 1 0 0

Other Southeast Asia 8 17 1 9 1 3 0 1 0 0 0 0

India 20 34 7 22 4 15 2 13 0 3 0 1

Other South Asia 7 11 6 8 2 2 0 1 1 1 0 0

Middle East 9 19 5 10 0 1 2 3 0 0 0 0

Africa 13 31 6 15 2 7 2 7 0 2 0 0

Oceania 2 5 0 2 0 1 0 0 0 0 0 0

Other 8 0 6 0 1 0 1 0 0 0 0 0 a Column (1) to (6) represent six size categories based on the number of single men within their own ethnicity and with a college or higher level of education at the MSA for a

given year. b Each cell in the table reports the number of MSAs fall into one size category for a given ethnicity and year. For example, for the US-born, the first column under the year 1990

shows that in 1990, there are 2 MSAs across the U.S. that have a MSA count of US-born col+ single men between 200 and 499.

49

Table A2-2: Distribution of Col+ Single Women among U.S. MSAs by Ethnicity (Birth Region):

Number of MSAs in 6 Size Categories a,

Size Categories Based on MSA Count of Own-Ethnic Co+ Single Women (W)

Year 1990 and Year 2010

(1) (2) (3) (4) (5) (6)

MSA Size Categories Are

Based on MSA Count of Own-

Ethnic Col+ Single Women (W)

200 ≤ W ≤ 499

500 ≤ W ≤ 999

1000 ≤ W ≤ 1999

2000 ≤ W ≤ 3999

4000 ≤ W ≤ 7999

W ≥ 8000

Year

1990

Year

2010

Year

1990

Year

2010

Year

1990

Year

2010

Year

1990

Year

2010

Year

1990

Year

2010

Year

1990

Year

2010

Ethnicity (Birth Region)

United States 0b 0 20 4 54 17 64 55 50 75 85 146

Canada 13 24 2 8 2 3 0 2 0 0 0 0

Other North America 0 2 0 0 0 0 0 0 0 0 0 0

Central America and Caribbean 15 35 11 24 4 14 1 11 1 5 2 4

Southern America 7 26 4 17 2 9 1 5 1 2 0 2

Northern Europe 10 25 7 6 2 4 1 1 0 1 0 0

Western Europe 18 19 4 9 2 4 1 2 0 1 0 0

Southern Europe 5 20 2 7 0 2 1 1 0 0 0 0

Eastern Europe 9 25 2 21 2 9 1 1 0 2 0 2

China 11 32 5 13 1 7 1 3 2 4 0 2

Japan 1 15 2 4 0 1 0 1 0 1 0 0

Korea 6 26 2 7 2 6 0 4 0 2 0 0

Other East Asia 0 2 0 0 0 0 0 0 0 0 0 0

Philippines 8 28 10 10 4 12 3 8 1 4 1 1

Vietnam 5 18 4 9 0 6 0 2 0 1 0 0

Other Southeast Asia 7 17 0 9 1 4 0 1 0 0 0 0

India 11 33 2 17 1 7 0 7 0 1 0 0

Other South Asia 6 13 1 3 1 1 0 2 0 0 0 0

Middle East 2 13 2 5 1 2 0 2 0 0 0 0

Africa 7 20 3 14 1 6 0 2 0 1 0 0

Oceania 2 5 0 4 0 2 0 0 0 0 0 0

Other 8 2 2 0 1 0 1 0 0 0 0 0 a Column (1) to (6) represent six size categories based on the number of single women with own ethnicity and with a college or higher level of education at the MSA for a given

year. b Each cell in the table reports the number of MSAs fall into one size category for a given ethnicity and year. For example, for the US-born, the first column under year 1990 shows

that in 1990, there is no MSA in the U.S that has a MSA count of US-born col+ single women falling between 200 and 499.

50

Table A3: Number of MSAs in 4 Size Categories Based on MSA Count of Own-Industry Ph.D. Workers (I)

PANEL A: Year 1990

MSA Size Categories Are Based on

MSA Count of Own-Industry Ph.D. Workers (I)

100≤ I ≤199 200≤ I ≤499

500≤ I ≤999

I ≥ 1000

Industry

Agriculture 9 2 0 0

Mining 2 2 1 1

Construction 13 7 2 0

Manufacturing 36 28 20 21

Transportation, Communication, Electricity and Gas 21 7 5 0

Wholesale 14 10 2 0

Retail 26 8 4 1

FIRE 28 12 8 3

Computer and Data Processing 13 8 2 0

Health Service 48 50 19 23

Legal Service 29 14 3 3

Education 41 66 48 90

Social Service 13 5 1 1

Museum, Art, Zoo 3 1 0 0

Membership Organization 30 17 3 3

Engineering 15 4 4 0

Accounting, Audit 2 1 0 0

R&D, Testing 21 22 12 12

Management and Public Relation 24 23 7 8

Other Services 20 10 3 1

Public Administration 33 31 10 3

PANEL B: Year 2010

MSA Size Categories Are Based on

MSA Count of Own-Industry Ph.D. Workers (I)

100≤ I ≤199 200≤ I ≤499

500≤ I ≤999

I ≥ 1000

Industry

Agriculture 27 17 2 0

Mining 5 2 0 1

Construction 14 5 2 0

Manufacturing 32 40 26 31

Transportation, Communication, Electricity and Gas 16 17 10 2

Wholesale 15 13 2 0

Retail 47 34 21 9

FIRE 29 25 12 7

Computer and Data Processing 14 13 11 5

Health Service 37 62 58 71

Legal Service 26 38 26 14

Education 41 66 48 90

Social Service 22 26 7 0

Museum, Art, Zoo 6 6 1 0

Membership Organization 33 49 13 8

Engineering 30 20 3 4

Accounting, Audit 7 5 0 0

R&D, Testing 26 41 17 22

Management and Public Relation 41 28 13 9

Other Services 23 19 10 3

Public Administration 38 50 23 10

51

Table A4-1: City Size Categories Based on MSA’s Own-Ethnic Population

Location Choices for Single Ph.D. Workers

City Size Categories Based on

MSA’s Own-Ethnic Population

Predicted Probabilities

Single Female Ph.D. Workers Single Male Ph.D. Workers Year 1990 Year 2000 Year 2010 Year 1990 Year 2000 Year 2010

Panel A: US-born

Large Own-Ethnic MSAs 0.3226 0.3208 0.2896 0.3155 0.3021 0.3104

(.00028) (.00008) (.00037) (.00011) (.00013) (.00038)

Midsize Own-Ethnic MSAs 0.3714 0.4236 0.4783 0.3991 0.4423 0.4633

(.00045) (.00026) (.00062) (.00039) (.00039) (.00105)

Small Own-Ethnic MSAs 0.3060 0.2556 0.2321 0.2854 0.2556 0.2263

(.00064) (.00031) (.00033) (.00037) (.00036) (.00075)

Panel B: Chinese

Large Own-Ethnic MSAs 0.5262 0.6054 0.4445 0.4842 0.4929 0.4607

(.00763) (.00483) (.01209) (.00703) (.00381) (.00873)

Midsize Own-Ethnic MSAs 0.2948 0.2774 0.3947 0.3097 0.3643 0.4371

(.00652) (.00203) (.01019) (.00476) (.00270) (.00647)

Small Own-Ethnic MSAs 0.1790 0.1172 0.1608 0.2061 0.1428 0.1022 (.00480) (.00320) (.00504) (.00549) (.00226) (.00283)

Panel C: Indian

Large Own-Ethnic MSAs 0.3507 0.4831 0.3869 0.3936 0.4706 0.3576

(.00572) (.00635) (.00643) (.00465) (.00295) (.00829)

Midsize Own-Ethnic MSAs 0.3639 0.3413 0.3784 0.3161 0.4084 0.5576

(.00788) (.00206) (.00220) (.00286) (.00267) (.00684)

Small Own-Ethnic MSAs 0.2854 0.1756 0.2347 0.2903 0.1210 0.0848

(.00696) (.00498) (.00677) (.00500) (.00158) (.00214) a City size categories are based on the size of the own-ethnic population that are aged 15 above and present in that MSA. For a given person, large own-ethnic MSAs are

metropolitan areas in which his/her own ethnic population rank within top 15 among all U.S. MSAs, midsize own-ethnic MSAs as those that rank outside top 15 but have at least

1,000 people within the ethnicity of this person (the cutoff number is 250,000 for the US-born) present in the local areas, and the rest of MSAs are classified as small own-ethnic

MSAs. b Within a specific gender-period, for each panel, the probabilities should sum to one. Standard errors are in parentheses.

52

Table A4-2: City Size Categories Based on MSA’s Own-Ethnic Population

Location Choices for Married Ph.D. Workers in Super-Power Couples (SPC)

City Size Categories Based on

MSA’s Own-Ethnic Population

Predicted Probabilities

Married Female Ph.D. Workers in SPC Married Male Ph.D. Workers in SPC Year 1990 Year 2000 Year 2010 Year 1990 Year 2000 Year 2010

Panel A: US-born

Large Own-Ethnic MSAs 0.3124 0.2960 0.2704 0.3222 0.2953 0.2740

(.00062) (.00035) (.00052) (.00029) (.00030) (.00085)

Midsize Own-Ethnic MSAs 0.3741 0.4258 0.4461 0.3642 0.4193 0.4442

(.00074) (.00055) (.00136) (.00060) (.00052) (.00114)

Small Own-Ethnic MSAs 0.3135 0.2782 0.2835 0.3136 0.2854 0.2818

(.00133) (.00086) (.00155) (.00076) (.00053) (.00097)

Panel B: Chinese

Large Own-Ethnic MSAs 0.4210 0.4841 0.4393 0.4267 0.4825 0.4444

(.00383) (.00214) (.00360) (.00909) (.00334) (.00351)

Midsize Own-Ethnic MSAs 0.3913 0.4104 0.4252 0.3704 0.4033 0.4334

(.00383) (.00107) (.00276) (.00396) (.00216) (.00308)

Small Own-Ethnic MSAs 0.1877 0.1055 0.1355 0.2029 0.1142 0.1222 (.00592) (.00132) (.00300) (.00683) (.00153) (.00196)

Panel C: Indian

Large Own-Ethnic MSAs 0.3534 0.4433 0.4328 0.3693 0.4590 0.4405

(.00445) (.00286) (.00570) (.00657) (.00407) (.00533)

Midsize Own-Ethnic MSAs 0.2754 0.4026 0.4273 0.3371 0.4108 0.4501

(.00349) (.00140) (.00651) (.00258) (.00250) (.00417)

Small Own-Ethnic MSAs 0.3712 0.1541 0.1399 0.2936 0.1302 0.1094

(.00668) (.00195) (.00507) (.00600) (.00216) (.00327) a City size categories are based on the size of the own-ethnic population that are aged 15 above and present in that MSA. For a given person, large own-ethnic MSAs are

metropolitan areas in which his/her own ethnic population rank within top 15 among all U.S. MSAs, midsize own-ethnic MSAs as those that rank outside top 15 but have at least

1,000 people within the ethnicity of this person (the cutoff number is 250,000 for the US-born) present in the local areas, and the rest of MSAs are classified as small own-ethnic

MSAs. b Within a specific gender-period, for each panel, the probabilities should sum to one. Standard errors are in parentheses.

53

Table A5-1: City Size Categories Based on MSA’s Total Population and Own-Ethnic Col+ Population (OECoAbv)

Location Choices for Single Ph.D. Workers a

City size categories Predicted Probabilities Single Female Ph.D. Workers Single Male Ph.D. Workers Year 1990 Year 2000 Year 2010 Year 1990 Year 2000 Year 2010

Panel A: US-born

Large MSAs + Large OECoAbv 0.2678 0.3063 0.2947 0.2645 0.2888 0.3313

(.00026) (.00011) (.00034) (.00013) (.00014) (.00029)

Large MSAs + Small OECoAbv 0.0000 0.0416 0.0779 0.0000 0.0426 0.0812

(.00000) (.00008) (.00011) (.00000) (.00008) (.00027)

Midsize MSAs + Large OECoAbv 0.0632 0.0288 0.0000 0.0506 0.0265 0.0000

(.00012) (.00006) (.00000) (.00007) (.00008) (.00000)

Midsize MSAs + Small OECoAbv 0.3724 0.3829 0.4077 0.4075 0.4024 0.3715

(.00038) (.00024) (.00050) (.00039) (.00035) (.00085)

Small MSAs 0.2966 0.2403 0.2198 0.2774 0.2397 0.2161

(.00064) (.00030) (.00041) (.00037) (.00035) (.00069)

Panel B: Chinese

Large MSAs + Large OECoAbv 0.3827 0.3922 0.3828 0.3109 0.3316 0.4147

(.00775) (.00357) (.01061) (.00380) (.00254) (.00782)

Large MSAs + Small OECoAbv 0.0053 0.0507 0.0883 0.0148 0.0631 0.0860

(.00060) (.00105) (.00338) (.00065) (.00093) (.00368)

Midsize MSAs + Large OECoAbv 0.1775 0.2000 0.0756 0.1735 0.1888 0.0898

(.00685) (.00318) (.00339) (.00374) (.00240) (.00253)

Midsize MSAs + Small OECoAbv 0.2542 0.2269 0.2629 0.2904 0.2517 0.2845

(.00422) (.00137) (.00614) (.00188) (.00133) (.00559)

Small MSAs 0.1803 0.1301 0.1904 0.2104 0.1647 0.1251 (.00840) (.00431) (.00861) (.00607) (.00371) (.00325)

Panel C: Indian

Large MSAs + Large OECoAbv 0.2951 0.3700 0.3540 0.2778 0.3530 0.3228

(.00692) (.00477) (.00512) (.00283) (.00232) (.00794)

Large MSAs + Small OECoAbv 0.0308 0.0292 0.0712 0.0323 0.0447 0.1150

(.00357) (.00051) (.00378) (.00128) (.00058) (.00403)

Midsize MSAs + Large OECoAbv 0.0709 0.0987 0.0714 0.1232 0.1310 0.0708

(.00323) (.00167) (.00427) (.00236) (.00156) (.00253)

Midsize MSAs + Small OECoAbv 0.3586 0.3384 0.2868 0.3149 0.3335 0.3778

(.00761) (.00134) (.00183) (.00120) (.00128) (.00578)

Small MSAs 0.2459 0.1637 0.2166 0.2514 0.1377 0.1135

(.01245) (.00575) (.01103) (.00568) (.00258) (.00237) a An MSA has large OECoAbv if its own-ethnic col+ population ranks within top15 among all U.S MSAs; otherwise it has Small OECoAbv. Standard errors are in parentheses.

54

Table A5-2: City Size Categories Based on MSA’s Total Population and Own-Ethnic Col+ Population (OECoAbv)

Location Choices for Married Ph.D. Workers in SPC

City size categories Predicted Probabilities Married Female Ph.D. Workers in SPC Married Male Ph.D. Workers in SPC Year 1990 Year 2000 Year 2010 Year 1990 Year 2000 Year 2010

Panel A: US-born

Large MSAs + Large OECoAbv 0.2576 0.2707 0.2812 0.2729 0.2727 0.2791

(.00056) (.00027) (.00062) (.00024) (.00034) (.00088)

Large MSAs + Small OECoAbv 0.0000 0.0301 0.0526 0.0000 0.0288 0.0537

(.00000) (.00007) (.00019) (.00000) (.00008) (.00027)

Midsize MSAs + Large OECoAbv 0.0565 0.0357 0.0000 0.0517 0.0331 0.0000

(.00015) (.00021) (.00000) (.00013) (.00016) (.00000)

Midsize MSAs + Small OECoAbv 0.3823 0.4042 0.3892 0.3724 0.3987 0.3945

(.00067) (.00052) (.00134) (.00056) (.00046) (.00103)

Small MSAs 0.3036 0.2593 0.2770 0.3030 0.2666 0.2727

(.00130) (.00083) (.00154) (.00075) (.00054) (.00090)

Panel B: Chinese

Large MSAs + Large OECoAbv 0.2509 0.3075 0.3605 0.2950 0.3131 0.3710

(.00365) (.00125) (.00354) (.00541) (.00190) (.00209)

Large MSAs + Small OECoAbv 0.0204 0.0691 0.1322 0.0149 0.0595 0.1224

(.00223) (.00075) (.00276) (.00083) (.00040) (.00227)

Midsize MSAs + Large OECoAbv 0.2208 0.1886 0.0754 0.1893 0.1884 0.0710

(.00647) (.00159) (.00121) (.00559) (.00134) (.00077)

Midsize MSAs + Small OECoAbv 0.3567 0.3151 0.2697 0.3225 0.3029 0.2684

(.00416) (.00173) (.00277) (.00616) (.00129) (.00179)

Small MSAs 0.1513 0.1197 0.1622 0.1783 0.1360 0.1672 (.00563) (.00171) (.00230) (.00645) (.00185) (.00185)

Panel C: Indian

Large MSAs + Large OECoAbv 0.2461 0.2857 0.3395 0.2688 0.2984 0.3533

(.00228) (.00220) (.00474) (.00446) (.00265) (.00442)

Large MSAs + Small OECoAbv 0.0211 0.0398 0.0852 0.0153 0.0318 0.0943

(.00154) (.00054) (.00248) (.00061) (.00028) (.00317)

Midsize MSAs + Large OECoAbv 0.1304 0.1307 0.0838 0.1221 0.1145 0.0759

(.00241) (.00148) (.00175) (.00297) (.00134) (.00113)

Midsize MSAs + Small OECoAbv 0.3703 0.3808 0.2960 0.3846 0.3784 0.3153

(.00325) (.00172) (.00516) (.00441) (.00175) (.00362)

Small MSAs 0.2320 0.1631 0.1955 0.2093 0.1469 0.1611

(.00602) (.00263) (.00371) (.00439) (.00266) (.00329) a An MSA has large OECoAbv if its own-ethnic col+ population ranks within top15 among all U.S MSAs; otherwise it has Small OECoAbv. Standard errors are in parenthese.

55

Table A6: Top 20 MSAs of Patent Creation in the Year 2001 a

Total

Patents at

Year 2001

Top Cited

Patents at

Year 2001b

Population at Year 1990

(In 1,000s)

College-Above Population

at Year 1990

(In 1,000s)

MSA Name All

ethnicities Chinese Indian Chinese Indian

San Jose, CA 4734 276 1173 32.81 13.25 14.46 6.96

San Diego, CA 1404 46 1951 9.15 3.11 3.30 1.14

Minneapolis-Saint

Paul, MN

1363 60 1756 3.59 3.90 1.75 1.70

Austin, TX 1024 47 551 3.01 2.60 1.30 1.05

Houston, TX 987 70 2376 15.72 17.32 6.75 7.62

Phoenix, AZ 784 49 1623 4.34 2.99 1.43 1.30

Atlanta, GA 776 36 1870 5.18 6.60 1.99 2.96

Rochester, NY 642 32 745 1.49 1.69 0.71 1.01

Cincinnati, OH-KY-

IN

465 23 964 1.07 2.48 0.45 1.36

Cleveland, OH 433 23 1366 2.82 3.87 0.89 2.12

Denver, CO 422 22 179 0.85 0.53 0.30 0.28

Milwaukee, WI 396 19 1197 2.78 1.04 0.89 0.55

Sacramento, CA 350 26 1094 1.67 1.39 0.86 0.66

Pittsburgh, PA 340 14 1128 11.39 3.36 2.22 0.97

Riverside-San

Bernardino, CA

330 14 1545 2.62 3.54 1.05 1.88

Raleigh-Durham, NC 310 14 1881 5.62 5.19 1.45 2.28

Saint Louis, MO-IL 290 7 1807 2.13 2.10 0.71 1.36

Indianapolis, IN 286 19 953 0.99 1.50 0.41 0.58

Baltimore, MD 281 11 1834 3.82 5.21 1.16 3.07

Ventura-Oxnard-

Simi Valley, CA

256 10 504 2.47 1.51 1.09 0.80

a These patents are utility patents with application year in 2001. Data are downloaded from NBER Patent Data Project (PDP) and

Harvard Patent Network Dataverse. Patents with multiple inventors are distributed evenly across all inventors’ corresponding

residing MSAs. Population count is calculated by using decennial census data from IPUMS. b Citation counts are until 2006, and are truncation-adjusted by using the weight Hall et al. (2001) suggested. Patents whose

number of received citations among top 10% within its own technology category are regarded as top cited patents.